Method and system for determining incidence relation between end face flatness and mechanical index of rock sample

Abstract

The invention provides a method and system for determining the incidence relation between the end face flatness and mechanical indexes of rock samples, and the method comprises the steps: preparing a plurality of rock samples with different end face flatness, collecting the multi-dimensional flatness parameters of the samples, measuring various mechanical indexes, water content and density of the samples, and generating a multi-dimensional data set; constructing a multi-dimensional input feature vector; constructing a correlation model by integrating a random forest model and a BP neural network, training the correlation model by taking a multi-dimensional data set as training data, and optimizing model hyper-parameters through cross validation; and constructing a differential correction model, and correcting the output of the correlation model through the differential correction model according to different lithology so as to obtain a mechanical index prediction curve corresponding to the multi-dimensional input feature vector of the rock sample to be tested. According to the method, the quantitative relation between the rock end face flatness and the mechanical index is established, and the influence of the end face flatness on the rock mechanical test can be effectively avoided.

Description

Technical Field

[0001] This application relates to the field of rock mechanics testing technology, and in particular to a method and system for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties. Background Technology

[0002] Currently, in rock mechanics testing, the flatness of the rock sample end face is a key factor affecting the test results. It is necessary to clarify the quantitative relationship between end face flatness and rock mechanics indicators in order to reduce the error in rock mechanics test results.

[0003] However, the relevant technologies only study the influence of joint roughness on rock strength, and the influence mechanism of end face smoothness on rock mechanical properties has not been systematically studied. Only the tolerance range of end face unevenness is specified, which leads to large dispersion of rock mechanical test results. The test results may contain errors due to the influence of end face smoothness.

[0004] Therefore, establishing a standardized correlation between rock end face flatness and mechanical properties has become an urgent problem to be solved. Summary of the Invention

[0005] This application aims to at least partially address one of the technical problems in the related art.

[0006] Therefore, the first objective of this application is to propose a method for determining the correlation between the flatness of the rock sample end face and its mechanical properties. This method quantifies the flatness parameters of the rock sample end face, constructs a correlation model and engineering database between flatness and mechanical properties, and establishes a quantitative relationship between the flatness of the rock end face and its mechanical properties. This method can effectively avoid the influence of end face flatness on rock mechanics testing and can be applied to various scenarios such as rock mechanics experiments and engineering rock mass stability assessment.

[0007] The second objective of this application is to propose a system for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties.

[0008] The third objective of this application is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, the first aspect of this application is to propose a method for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties, comprising the following steps:

[0010] Multiple rock samples with different end-face flatness were prepared using a flatness preparation process. Multidimensional flatness parameters of the multiple rock samples were collected, and various mechanical properties, water content and density of the multiple rock samples were measured to generate a multidimensional dataset.

[0011] Based on the smoothness parameters, lithology, water content, and density, a multidimensional input feature vector is constructed;

[0012] An association model is constructed by integrating a random forest model and a backpropagation neural network. The multidimensional dataset is used as training data to train the association model. The model hyperparameters are optimized by cross-validation to dynamically adjust the feature weights for lithology. The association model uses the multidimensional input feature vector as input to predict the mechanical indices corresponding to the input features.

[0013] A differential correction model is constructed to correct the output of the correlation model for different lithologies. Based on the differential correction model, the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested is obtained.

[0014] Optionally, in one embodiment of this application, the association model is specifically used to: identify the interaction effects between different parameters in the multidimensional input feature vector using the Multivariate Adaptive Regression Spline (MARS) algorithm.

[0015] Optionally, in one embodiment of this application, the dynamic adjustment of feature weights for lithology includes: setting weight adjustment factors for water content and density for lithology, wherein when the rock sample is soft rock, the weight of water content is increased, and when the rock sample is hard rock, the weight of density is increased; and setting weight thresholds for water content and density in the random forest model.

[0016] Optionally, in one embodiment of this application, the step of modifying the output of the correlation model using the differentiated correction model for different lithologies includes: stratifying the multidimensional input feature vector into data layers and standardizing the layered data; for soft rock, constructing a first correction formula with uniaxial compressive strength as the dependent variable and standardized roughness parameters, water content, and density as independent variables, wherein the first correction formula contains multiple lithology correlation coefficients to be solved; performing ridge regression fitting on the first correction formula to construct a loss function containing an L2 regularization term; and optimizing the regularization parameter in the loss function through 5-fold cross-validation to obtain multiple fitted lithology correlation coefficients.

[0017] Optionally, in one embodiment of this application, after standardizing the stratified data, the method further includes: for hard rock, selecting multiple target features sensitive to hard rock from the standardized data, and constructing a second correction formula with uniaxial compressive strength as the dependent variable and the multiple target features as independent variables, wherein the second correction formula includes multiple lithological correlation coefficients to be solved; performing least squares fitting on the second correction formula and constructing a three-dimensional particle flow model (PFC); performing discrete element simulation using the three-dimensional particle flow model PFC to calibrate the lithological correlation coefficients corresponding to density; expanding the data in the training dataset based on the discrete element simulation, and correcting the correlation model using the expanded training dataset.

[0018] Optionally, in one embodiment of this application, obtaining the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested includes: receiving the multidimensional input feature vector of the rock sample to be tested through a visualization query platform; calling the trained prediction model to output the mechanical index prediction value corresponding to the currently received multidimensional input feature vector, wherein the prediction model includes the association model and the differential correction model; within a preset flatness range, using the mechanical index prediction value to generate equally spaced data points according to unit step size, and obtaining the mechanical index prediction curve through difference fitting.

[0019] Optionally, in one embodiment of this application, after obtaining the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested, the method further includes: displaying the mechanical index prediction curve through the interactive interface of the visualization query platform; and generating a strength correction coefficient table within the preset flatness range.

[0020] Optionally, in one embodiment of this application, the preparation of multiple rock samples with different end-face flatness using the flatness preparation process includes: combining high-precision grinding and polishing equipment and pressure sensors to adjust the end-face flatness of different rock samples.

[0021] To achieve the above objectives, a second aspect of this application also proposes a system for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties, comprising the following modules:

[0022] The acquisition module is used to prepare multiple rock samples with different end face flatness using a flatness preparation process, acquire multidimensional flatness parameters of the multiple rock samples, and measure various mechanical properties, water content and density of the multiple rock samples to generate a multidimensional dataset.

[0023] A construction module is used to construct a multidimensional input feature vector based on the smoothness parameters, lithology, water content, and density;

[0024] The prediction module is used to construct an association model by integrating a random forest model and a backpropagation neural network, train the association model using the multidimensional dataset as training data, and optimize the model hyperparameters through cross-validation to dynamically adjust the feature weights for lithology. The association model uses the multidimensional input feature vector as input to predict the mechanical indicators corresponding to the input features.

[0025] The correction module is used to construct a differential correction model, which corrects the output of the correlation model for different lithologies, and obtains the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested based on the differential correction model.

[0026] To implement the above embodiments, a third aspect of this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for determining the correlation between the flatness of the rock sample end face and mechanical properties as described in the first aspect.

[0027] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects: This application constructs a multi-dimensional feature vector combining three-dimensional morphology and physical properties in rock mechanics testing, improves the feature system, and combines machine learning with experimental data to predict correlations. Compared with traditional statistical regression models, it has higher prediction accuracy, achieving a leap from qualitative control of smoothness parameters to quantitative prediction of mechanical indicators. Furthermore, this application introduces dynamic weights, cross-domain feature fusion, and interpretability analysis into the prediction model, improving the adaptability of the machine learning model in geotechnical engineering. It also introduces a dynamic adjustment mechanism for the importance of lithology-sensitive features into the random forest algorithm, which, compared to general machine learning frameworks, better aligns with the objective laws of the sensitivity of lithology-dominant parameters in rock mechanics. Moreover, this application constructs an engineering database that can be directly applied to rock mechanics testing laboratories and engineering rock mass stability assessment scenarios, reducing repetitive experiments and lowering testing costs caused by smoothness errors. Therefore, this application establishes a quantitative relationship between rock end-face smoothness and mechanical indicators, effectively avoiding the influence of end-face smoothness on rock mechanics testing, improving the accuracy of quantitative relationship prediction, and contributing to improved engineering safety and economy.

[0028] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0029] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0030] Figure 1This is a flowchart illustrating a method for determining the correlation between the end face flatness and mechanical properties of a rock sample, as proposed in an embodiment of this application.

[0031] Figure 2 This is a flowchart of a differentiation correction method proposed in an embodiment of this application;

[0032] Figure 3 A flowchart of another differentiation correction method proposed in the embodiments of this application;

[0033] Figure 4 A flowchart illustrating a method for generating and displaying a mechanical index prediction curve according to an embodiment of this application;

[0034] Figure 5 This is a schematic diagram of the structure of a system for determining the correlation between the end face flatness and mechanical properties of a rock sample, as proposed in an embodiment of this application. Detailed Implementation

[0035] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0036] The following describes, with reference to the accompanying drawings, a method and system for determining the correlation between the end face flatness and mechanical properties of a rock sample, as proposed in an embodiment of this application.

[0037] Figure 1 This is a flowchart illustrating a method for determining the correlation between the end-face flatness and mechanical properties of a rock sample, as proposed in an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0038] Step S101: Prepare multiple rock samples with different end face flatness using a flatness preparation process, collect multidimensional flatness parameters of multiple rock samples, and measure multiple mechanical properties, water content and density of multiple rock samples to generate a multidimensional dataset.

[0039] Specifically, this application first uses a flatness preparation process to prepare rock samples with different end face flatness.

[0040] In one embodiment of this application, a flatness preparation process is used to prepare multiple rock samples with different end face flatness, including: combining high-precision grinding and polishing equipment and pressure sensors to adjust the end face flatness of different rock samples.

[0041] Specifically, high-precision grinding and polishing equipment (e.g., CNC grinding machines) is used to grind and polish the sample end faces. A pressure sensor (monitoring accuracy ±0.1N) is used to monitor the pressure applied by the grinding and polishing equipment in real time. Furthermore, by controlling process parameters such as grinding and polishing time (e.g., 5-30 min) and grinding wheel mesh size (e.g., 80-1000 mesh), rock samples with different degrees of flatness can be prepared. Therefore, the flatness error of the sample end face prepared in this embodiment can be less than or equal to ±0.01 mm.

[0042] Furthermore, multidimensional smoothness parameters of multiple prepared rock samples are collected through multidimensional detection methods. For example, a three-dimensional laser scanner (accuracy ±5μm) can be used to acquire the end face morphology of the samples, and then roughness parameters (including arithmetic mean roughness Ra and maximum height roughness Rz), fractal dimension D, and root mean square deviation σ, etc., are calculated based on the scanning data. This allows the establishment of a basic dataset containing more than 5000 samples.

[0043] Furthermore, by conducting corresponding mechanical experiments on rock samples of different lithologies, various mechanical properties of multiple rock samples can be determined. For example, for different rock types (such as sandstone, limestone, and shale), uniaxial compression tests (loading rate 0.05-2 mm / min), triaxial compression tests (confining pressure 0-100 MPa), and Brazilian splitting tests can be performed, simultaneously recording various mechanical properties such as peak strength, elastic modulus, and Poisson's ratio for each rock sample. Combined with the established smoothness parameters of each rock sample, a table of correspondence between smoothness parameters and mechanical properties can be generated. The moisture content (w) of each rock sample can also be determined using the drying method, and the density (ρ) of each rock sample can be calculated using the volumetric method.

[0044] Therefore, this application obtains a multidimensional dataset of rock samples through various methods such as field measurements and experiments.

[0045] Step S102: Construct a multidimensional input feature vector based on flatness parameters, lithology, water content, and density.

[0046] Specifically, for the obtained multidimensional dataset, feature engineering is performed to construct a multidimensional input feature vector containing multiple smoothness parameters, lithology, water content, and density, in order to determine the input features of the prediction model.

[0047] For example, multiple smoothness parameters (Ra, Rz, D, and σ) are correlated with lithology (i.e., rock type, which can include sandstone, limestone, and shale, etc.), water content (0-20%, divided into intervals of 5%), and density (2.0-3.0 g / cm³). 3 Accurate to 0.01 g / cm³ 3Independent thermal encoding and normalization are performed to construct a 10-dimensional input feature vector. For example, it can be in the form of 3-dimensional lithology independent thermal encoding + 4-dimensional flatness + 1-dimensional water content + 1-dimensional density + 1-dimensional other data.

[0048] In this application, water content and density are used as independent variables in the training of the mechanical index prediction model. Unlike traditional materials science models that only use a single roughness parameter (such as Ra in the field of machining), this application constructs a multidimensional feature system that includes three-dimensional morphology parameters (fractal dimension D and root mean square deviation σ) and physical property parameters (water content and density), thus overcoming the limitation of using a single parameter to characterize surface integrity.

[0049] For example, in mechanical engineering, gear surface roughness models typically only contain Ra and machining parameters, while this application can also characterize the self-similarity of microcracks on rock surfaces through fractal dimension D, and correlate mineral particle cementation strength through density parameters to form a feature space suitable for multi-scale failure mechanisms of rocks.

[0050] Therefore, this application addresses the difficulty in quantifying the complex surface of rock samples by constructing a multi-parameter evaluation system to replace a single roughness parameter, which can accurately characterize the three-dimensional morphology of rock samples.

[0051] Step S103: Construct an association model by integrating a random forest model and a backpropagation neural network. Use a multidimensional dataset as training data to train the association model and optimize the model hyperparameters through cross-validation to dynamically adjust the feature weights for lithology.

[0052] The correlation model uses a multi-dimensional input feature vector as input to predict the mechanical indices corresponding to the input features.

[0053] Specifically, this application constructs a machine learning model for predicting mechanical indicators, employing an integrated modeling approach using a random forest (RF) algorithm and a backpropagation neural network (BP-NN). The multi-dimensional input feature vectors determined in the previous step are used to design the model's input features, which include lithology, water content, density, and smoothness parameters. Furthermore, hyperparameters are optimized through cross-validation; for example, the random forest tree depth is set to 20-50, and the number of hidden layer nodes in the BP neural network is set to 10-30. Feature weights are dynamically adjusted through feature cross-validation.

[0054] In one embodiment of this application, the constructed association model is specifically used to: identify the interaction effects between different parameters in a multidimensional input feature vector using the Multivariate Adaptive Regression Spline (MARS) algorithm.

[0055] Specifically, since this application incorporates moisture content, density, and flatness parameters into a unified feature space, it models the interaction of multiple physical parameters and uses the Multivariate Adaptive Regression Splines (MARS) algorithm to identify the interaction effects between various parameters, such as the synergistic attenuation effect of the product of moisture content and Ra on strength.

[0056] Therefore, this application can perform multi-factor coupling analysis, thereby improving the accuracy of prediction.

[0057] In one embodiment of this application, the feature weights of lithology are dynamically adjusted, including: setting weight adjustment factors for water content and density for lithology, wherein when the rock sample is soft rock, the weight of water content is increased, and when the rock sample is hard rock, the weight of density is increased; and setting weight thresholds for water content and density in the random forest model.

[0058] Specifically, this embodiment adds weight adjustment factors for water content and density when optimizing hyperparameters through cross-validation. As a possible implementation, a lithology-sensitive dynamic adjustment mechanism for feature importance is introduced into the random forest algorithm; for example, the weight of water content is increased to 30% in soft rock and the weight of density is increased to 25% in hard rock. Furthermore, this embodiment sets water content feature importance thresholds of ≥0.15 and density of ≥0.1 in the RF algorithm, defining a basic range for the dynamic adjustment of weights.

[0059] Therefore, unlike traditional models that treat all features with equal weight, this embodiment is more in line with the objective law of sensitivity of lithological dominant parameters, and forces the model to learn the coupling effect of water content, density and smoothness, which can improve the accuracy of prediction.

[0060] Based on the above model construction method, this application uses the multidimensional dataset obtained in step S101 to generate the training dataset required for model training, and trains the above-mentioned related model.

[0061] As an example, a training dataset containing 100,000 experimental data points was constructed by augmenting the data. Stratified sampling was used based on water content (low, medium, high) and density (light, medium, heavy). Soft rock (e.g., shale) samples comprised 30% (with each water content sample comprising 10%), and hard rock (e.g., granite) comprised 20% (with each density sample comprising 5%-8%). Cross-validation was performed 10 times. Furthermore, model evaluation metrics were defined, including R... 2 Root mean square error (RMSE) and mean absolute error (MAE), for example, setting the model prediction accuracy R... 2The root mean square error (RMSE) is ≤3% and ≥0.92. The specific model training process can refer to the training methods in related technologies; this application does not impose any restrictions on it.

[0062] It should be noted that this application transforms the coupling problem of lithology, smoothness, and environmental parameters in rock mechanics into a nonlinear mapping in a high-dimensional feature space. Compared to concrete strength prediction models in civil engineering (such as those typically employing linear regression or a single neural network), this application achieves feature cross-validation by setting weight adjustment factors for moisture content and density. Furthermore, drawing on the multi-scale feature extraction concept from image processing, it achieves cross-domain model fusion. Specifically, it captures the nonlinear influence of smoothness parameters through the tree structure of a random forest, and combines this with a BP neural network to fit the complex coupling relationship between moisture content, density, and strength, forming a cross-domain correlation model of "geometric features - physical properties - mechanical response."

[0063] Step S104: Construct a differential correction model. For different lithologies, the output of the correlation model is corrected by the differential correction model. Based on the differential correction model, the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested is obtained.

[0064] Specifically, since there are various types of rock samples in practical applications, and lithological differences can lead to model prediction biases, this application also constructs a differential correction model. This model corrects the output of the aforementioned correlation model by generating a multivariate correction formula for mechanical indices. Thus, the prediction model of this application, composed of the correlation model and the differential correction model, can more accurately predict the correlation between the end-face smoothness of rock samples and mechanical indices.

[0065] Among them, the differentiated correction model of this application can establish a multivariate correction formula that includes water content and density for soft rocks (such as shale) and hard rocks (such as granite).

[0066] To more clearly illustrate the specific implementation of the differentiated modification of lithology in this application, the following description uses a modification method proposed in one embodiment of the application as an example. Figure 2 This is a flowchart of a differentiation correction method proposed in an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0067] Step S201: The multidimensional input feature vector is divided into data layers, and the layered data is standardized.

[0068] Specifically, the data is stratified according to lithology (soft or hard rock), water content (0-20%, with 5% as a increment), and density (0.1 g / cm³). 3Divide the data into subsets at intervals to ensure that each subset contains at least 500 samples.

[0069] Then, standardization is performed by Z-score standardization of the flatness parameters (Ra and Rz), moisture content (w), and density (ρ) to eliminate the influence of dimensions.

[0070] Step S202: For soft rock, construct a first correction formula with uniaxial compressive strength as the dependent variable and standardized roughness parameters, water content and density as independent variables. The first correction formula contains multiple lithological correlation coefficients to be solved.

[0071] Specifically, for soft rock, using the measured uniaxial compressive strength (UCS) as the dependent variable and Ra, water content (w), and density (ρ) as independent variables, the following first correction formula is constructed:

[0072] UCS=α×Ra+β×w+γ×ρ+δ

[0073] Where α, β, γ and δ are lithological correlation coefficients.

[0074] Step S203: Ridge regression is performed on the first modified formula to construct a loss function that includes an L2 regularization term.

[0075] Specifically, the lithological correlation coefficients mentioned above are solved using ridge regression fitting. Ridge regression fitting is a regularized improvement on the least squares method. When performing ridge regression fitting on the first modified formula, a loss function with an L2 regularization term is constructed as shown in the following formula:

[0076]

[0077] Among them, y i Let λ be the measured intensity of the i-th sample, and λ be the regularization parameter.

[0078] Step S204: Optimize the regularization parameter in the loss function through 5-fold cross-validation to obtain multiple lithological correlation coefficients obtained from the fitting.

[0079] Specifically, 5-fold cross-validation is a model evaluation method used to assess the performance and generalization ability of machine learning models. This embodiment optimizes the regularization parameter λ (search range 0.01-10) in the loss function using 5-fold cross-validation, thereby suppressing possible multicollinearity between water content and density (e.g., a positive correlation between water content and density in clay-rich soft rock), and improving the validation set R. 2 ≥0.92.

[0080] Therefore, this embodiment uses ridge regression to fit the measured data, introduces L2 regularization to suppress overfitting, and then optimizes the regularization parameters through 5-fold cross-validation to determine multiple correlation coefficients in step S202 above. This embodiment can also verify the accuracy of the solved correlation coefficients through sensitivity analysis. For example, if the solved β < 0, it conforms to the rule that higher moisture content results in lower strength.

[0081] Based on the above embodiments, this application can also perform differentiated modifications on hard rock. Figure 3 A flowchart of another differentiation correction method proposed in the embodiments of this application is shown below. Figure 3 As shown, after step S201 is executed in the above embodiment, the method further includes:

[0082] Step S301: For hard rock, select multiple target features that are sensitive to hard rock from the standardized data, and construct a second correction formula with uniaxial compressive strength as the dependent variable and multiple target features as independent variables. The second correction formula contains multiple lithological correlation coefficients to be solved.

[0083] Specifically, for hard rock (such as granite), significant variables (such as Ra and ρ) are screened using stepwise regression to eliminate parameters that are insensitive to hard rock, such as water content. Then, the second corrected formula is fitted using the least squares method, as shown in the following formula: UCS=α×Ra+γ×ρ+ε.

[0084] Step S302: The second modified formula is fitted with the least squares method, and a three-dimensional particle flow model PFC is constructed. Discrete element simulation is performed through the three-dimensional particle flow model PFC to calibrate the lithological correlation coefficient corresponding to the density.

[0085] Specifically, the coefficient γ corresponding to density ρ in the second modified formula can be calibrated using discrete element method (DEM) simulation. Specifically, the Particle Flow Code (PFC) model can be used, requiring the residual standard deviation to be ≤3% of the measured value. The PFC model is a numerical simulation tool based on the discrete element method (DEM) that predicts the macroscopic behavior of materials by simulating the interactions between particles.

[0086] It should be noted that because the cementation of mineral particles in hard rocks is significant, their strength is mainly controlled by the contact stiffness between particles and the cementation strength. Therefore, discrete element method (DEM) simulations can intuitively characterize the microcrack propagation process through particle flow model (PFC). However, for soft rocks (such as shale), due to the complex water absorption and softening characteristics of clay minerals, DEM cannot accurately characterize the water-rock interaction of nanoscale clay particles, and therefore requires the reliance on measured data.

[0087] Therefore, this embodiment uses the least squares method to fit the measured data and the discrete element virtual data, and then uses the discrete element simulation calibration coefficient γ to determine the multiple correlation coefficients in step S301 above. This embodiment can also verify the accuracy of the solved correlation coefficients through residual analysis, for example, by determining whether the standard deviation of the residuals of the simulation results is less than or equal to 3% of the measured value.

[0088] Step S303: Expand the data in the training dataset based on discrete element simulation, and correct the correlation model using the expanded training dataset.

[0089] Specifically, in this embodiment, when making prediction corrections for hard rock, discrete element simulation can be used to expand the data in the training dataset, thereby using the expanded training dataset to further optimize the parameters in the above-mentioned correlation model, so as to improve the prediction accuracy for hard rock.

[0090] As one possible implementation, the process of augmenting the training dataset based on discrete element simulation includes the following steps:

[0091] The first step is model construction. Based on the mineral composition of hard rock (such as quartz and feldspar), a particle aggregate model is established, and the particle size distribution (1-5 mm), contact model (linear parallel cementation model), and cementation parameters (normal or tangential stiffness and cementation strength) are set.

[0092] The second step is parameter calibration. Through simulated uniaxial compression experiments, the bonding parameters are adjusted to ensure the error between the simulated strength and the measured value is ≤5%, establishing a quantitative relationship between density (ρ) and bonding strength (e.g., for every 0.1 g / cm³ increase in ρ). 3 The bonding strength is increased by 3-6 MPa.

[0093] The third step is to generate data. Within the standard flatness range (Ra≤50μm), the Ra parameter is simulated by changing the surface roughness of the particle contact surface, generating 20,000 virtual samples (containing the correspondence between Ra, ρ and UCS). These samples are then mixed into the original training set at a 1:1 ratio to improve the model's generalization ability for multi-scale damage in hard rock.

[0094] It should be noted that the impact of cross-scale fusion of lithology, water content, and density in this embodiment can be corrected for model biases caused by lithological differences through stratified sampling and data augmentation. Specifically, for soft rocks, experimental samples can be added by 20% according to water content stratification; for example, 500 new data points can be added to each of the 10%, 15%, and 20% water content groups in the shale samples. For hard rocks, discrete element simulation is used to supplement the data.

[0095] It should also be noted that when performing sensitivity analysis in the above embodiments, the feature importance ranking of random forests can be used (e.g., w weight 30%, Ra weight 25% in soft rock) to verify the consistency between the coefficient sign and physical meaning (e.g., β<0 indicates that the higher the water content, the lower the strength). When fitting and solving for the correlation coefficient, parameter optimization can be performed, for example, by using gradient descent or matrix inversion (ridge trace method) to find the coefficients that minimize the loss function. Here, β<0 indicates that for every 1% increase in water content, the strength decreases by 0.5-1.2 MPa, and γ>0 indicates that for every 0.1 g / cm³ increase in density... 3 Strength increases by 2-5 MPa; when Ra > 10 μm and water content > 15%, the strength attenuation rate of soft rock increases to 20%-25%; when density > 2.6 g / cm³ 3 At that time, the sensitivity of hard rock to flatness decreased by 30%.

[0096] Based on the above embodiments, the prediction model constructed in this application, which consists of an association model and a differentiation correction model, can also be validated. For example, the fitting accuracy can be evaluated through 5-fold cross-validation, requiring the coefficient of determination R between the predicted value and the measured value to be... 2 The root mean square error (RMSE) is ≤5%, with an accuracy ≥0.92. Furthermore, it is understood that, compared to ordinary least squares (OLS) without regularization, the ridge regression in this application, by introducing a weight decay term, can effectively address the multicollinearity problem that may exist in rock physical parameters (such as water content and density), thereby improving the model's generalization ability. For example, when the water content and density of soft rock are positively correlated, ridge regression can avoid the problem of excessively large variance in coefficient estimation caused by strong correlations between parameters in OLS.

[0097] Therefore, the prediction model established in this application can accurately predict the mechanical indices corresponding to different multidimensional input feature vectors.

[0098] Furthermore, after training and adjusting the prediction model, in practical applications, the multidimensional input feature vector of the rock sample to be tested can be input into the prediction model to obtain the predicted values ​​of multiple mechanical indicators output by the prediction model, and then the corresponding mechanical indicator prediction curve can be generated based on the predicted values.

[0099] To more clearly illustrate the specific implementation of generating and displaying the mechanical index prediction curve in this application, the following example illustrates a method proposed in one embodiment of this application. Figure 4 This is a flowchart illustrating a method for generating and displaying a mechanical index prediction curve according to an embodiment of this application, as shown below. Figure 4 As shown, the method includes the following steps:

[0100] Step S401: Receive the multidimensional input feature vector of the rock sample to be tested through the visualization query platform.

[0101] Specifically, this embodiment pre-constructs a visual query platform, which includes a web-based interactive interface that allows users to input the smoothness parameters, lithology, and water content of the rock sample to be tested. The platform can then output the intensity range and prediction curve in real time based on the input parameters.

[0102] The prediction curve in this application is a quantitative relationship curve, where the horizontal axis represents the smoothness parameters (Ra and Rz, etc.) and the vertical axis represents the mechanical properties (such as uniaxial compressive strength UCS and elastic modulus E). Under the condition of fixed lithology, water content and density, it can reflect the quantitative influence trend of smoothness parameters on mechanical properties.

[0103] Step S402: Call the trained prediction model to output the predicted mechanical index value corresponding to the currently received multi-dimensional input feature vector.

[0104] The prediction model is the model that has been trained and adjusted as described above, including the correlation model and the differential correction model.

[0105] Specifically, when the visualization query platform processes the input data, it operates on a data-driven basis. That is, the platform calls a pre-trained prediction model to extract the feature weights and regression equations corresponding to the input parameters. For example, for a certain sandstone, the regression equation is obtained as shown in the formula: UCS = -3.2 × Ra - 1.5 × w + 4 × ρ + 50. Then, the platform can output the predicted mechanical properties of that sandstone.

[0106] Step S403: Within the preset flatness range, generate equally spaced data points using the predicted mechanical index values ​​at unit step lengths, and obtain the predicted mechanical index curve through difference fitting.

[0107] Specifically, when generating the curve, the lithology, water content, and density parameters are fixed, and within the flatness range allowed by the national standard (Ra≤50μm), equally spaced data points are generated with a step size of 5μm. A continuous curve is obtained by fitting the curve through cubic spline interpolation.

[0108] Step S404: Display the mechanical index prediction curve through the interactive interface of the visualization query platform.

[0109] Specifically, after obtaining the predicted mechanical index curves corresponding to the multidimensional input feature vectors of the rock sample to be tested, a front-end display can be performed. As an example, a two-dimensional "smoothness-strength" curve is dynamically plotted using visualization libraries such as ECharts and displayed on the interactive interface of the visualization query platform. The horizontal axis of this curve represents Ra or Rz, and the vertical axis represents the predicted mechanical index value. Simultaneously, the scatter points of the measured data (error ±5%) and the model determination coefficient R are labeled. 2 .

[0110] Furthermore, after obtaining the mechanical index prediction curves corresponding to the multidimensional input feature vectors of the rock sample to be tested, this embodiment also includes: generating a strength correction coefficient table within a preset flatness range. Specifically, this embodiment can also construct a standard-compatible module based on a large number of prediction results, establish a strength correction formula within the flatness range allowed by national sample preparation standards (e.g., Ra≤50μm), form a standardized prediction system, and automatically generate a standard-compliant strength correction coefficient table, thereby supporting one-click generation of engineering reports.

[0111] Based on the above embodiments, this application can also establish an engineering database system for rock mechanical indices. The database architecture can adopt a relational database (MySQL) and be able to store experimental data (≥100,000 records), model parameters (≥200 sets), and engineering cases (≥50 cases). The designed fields include sample ID, lithology, smoothness parameters, mechanical indices, and experimental conditions.

[0112] As an example, this database system includes the following modules:

[0113] The data acquisition module is used to store experimental data, numerical simulation results, and engineering cases.

[0114] The model prediction module integrates the prediction model described in the above embodiments, supports inputting lithology, water content, density and smoothness parameters, and supports outputting the strength range using multivariate correction formulas.

[0115] The visualization and interaction module provides functions such as displaying the flatness-strength relationship curve, data export, and API interface.

[0116] This example's model prediction module incorporates a lithological differentiation correction formula, and its database covers the smoothness range allowed by national sample preparation standards (Ra≤50μm), along with a built-in lithological differentiation correction model. When applied to engineering rock mass stability assessments, this database system can output a strength range conforming to national standards by inputting the smoothness parameters of the rock sample's end face, guiding engineering design.

[0117] In summary, the method for determining the correlation between the end-face smoothness of rock samples and mechanical indicators in this application constructs a multi-dimensional feature vector combining three-dimensional morphology and physical properties in rock mechanics testing, improves the feature system, and combines machine learning with experimental data to predict the correlation. Compared with traditional statistical regression models, it has higher prediction accuracy, achieving a leap from qualitative control of smoothness parameters to quantitative prediction of mechanical indicators. Furthermore, this method introduces dynamic weights, cross-domain feature fusion, and interpretability analysis into the prediction model, improving the adaptability of the machine learning model in geotechnical engineering. It also introduces a dynamic adjustment mechanism for the importance of lithology-sensitive features into the random forest algorithm, which, compared to general machine learning frameworks, better aligns with the objective laws of the sensitivity of lithology-dominant parameters in rock mechanics. Moreover, this method constructs an engineering database that can be directly applied to rock mechanics testing laboratories and engineering rock mass stability assessment scenarios, reducing repetitive experiments and lowering testing costs caused by smoothness errors. Therefore, this method establishes a quantitative relationship between rock end face flatness and mechanical properties, which can effectively avoid the influence of end face flatness on rock mechanical testing, improve the accuracy of quantitative relationship prediction, and help improve engineering safety and economy.

[0118] Based on the above embodiments, in order to more clearly and intuitively describe the method for determining the correlation between the flatness of the rock sample end face and the mechanical index of this application, the specific process and the effect achieved in practical applications are illustrated below with several specific embodiments in practical applications.

[0119] The first embodiment predicts the uniaxial compressive strength of samples from different lithologies. This embodiment includes the following steps:

[0120] The first step is sample preparation. Quartz sandstone (density 2.3 g / cm³) was selected. 3 Three groups of samples with moisture contents of 5%, 10%, and 15% were prepared. Ten samples with Ra = 10μm, 5μm, and 2μm were prepared from each group using 80-mesh, 200-mesh, and 500-mesh grinding wheels.

[0121] The second step was data collection. The UCS of the sample with 5% moisture content and Ra=10μm was measured to be 52MPa, and the UCS of the sample with 15% moisture content and Ra=10μm was 45MPa (the strength decreases by approximately 7-8MPa for every 5% increase in moisture content); the density was 2.5g / cm³. 3 The granite sample (Ra = 5 μm) had a UCS of 125 MPa and a density of 2.7 g / cm³. 3 The UCS of the same lithological sample was 132 MPa (for every 0.2 g / cm³ increase in density). 3 The strength is increased by approximately 7 MPa.

[0122] The third step is model training. The data is input into the random forest model, and the sandstone correction formulas for water content (w) and density (ρ) are obtained through training, as shown below: UCS = -3.2 × Ra - 1.5 × w + 4 × ρ + 50(R) 2 =0.97), the granite correction formula is as follows: UCS=-2×Ra+3×ρ+55(R 2 =0.95).

[0123] The fourth step is engineering application. Sandstone samples were collected from a slope engineering project, and the measured Ra was 3 μm. A database query yielded a UCS predicted value of 61.6 MPa, while the field test value was 60 MPa, confirming an error of ≤2.7%.

[0124] The second embodiment involves operating the constructed database platform. This embodiment includes the following steps:

[0125] The first step is to obtain user input data: select shale, water content 15%, and Ra=15μm in the web interface.

[0126] The second step is system processing. The differential model is invoked, outputting a uniaxial compressive strength range of 25-30 MPa, while simultaneously generating a strength-flatness relationship curve and an experimental data traceability report.

[0127] The third step is to verify the results by comparing the experimental data of the three groups of shale samples and determining that the predicted value and the measured value have a consistency of ≥92%.

[0128] To achieve the above embodiments, this application also proposes a system for determining the correlation between the end-face flatness and mechanical properties of rock samples. Figure 5 This is a schematic diagram of the structure of a system for determining the correlation between the end face flatness and mechanical properties of a rock sample, as proposed in an embodiment of this application. Figure 5 As shown, the system includes: a data acquisition module 100, a data construction module 200, a prediction module 300, and a correction module 400.

[0129] The acquisition module 100 is used to prepare multiple rock samples with different end face flatness using a flatness preparation process, acquire multidimensional flatness parameters of multiple rock samples, and measure multiple mechanical properties, water content and density of multiple rock samples to generate a multidimensional dataset.

[0130] Module 200 is used to construct a multidimensional input feature vector based on flatness parameters, lithology, water content, and density.

[0131] The prediction module 300 is used to construct an association model by integrating a random forest model and a BP neural network. The association model is trained using a multidimensional dataset as training data, and the model hyperparameters are optimized through cross-validation to dynamically adjust the feature weights for lithology. The association model takes a multidimensional input feature vector as input to predict the mechanical indices corresponding to the input features.

[0132] The correction module 400 is used to construct a differential correction model. It corrects the output of the associated model for different lithologies and obtains the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested based on the differential correction model.

[0133] It should be noted that the explanation of the aforementioned method for determining the correlation between the flatness of the rock sample end face and its mechanical properties also applies to the system in this embodiment, and will not be repeated here.

[0134] In summary, the system for determining the correlation between rock sample end face flatness and mechanical properties in this application establishes a quantitative relationship between rock end face flatness and mechanical properties, effectively avoiding the influence of end face flatness on rock mechanical testing and improving the accuracy of quantitative relationship prediction.

[0135] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties as described in any one of the first aspects of the embodiments above.

[0136] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0137] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0138] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0139] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0140] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0141] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0142] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0143] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for determining the correlation between the flatness of the end face of a rock sample and its mechanical properties, characterized in that, Includes the following steps: Multiple rock samples with different end-face flatness were prepared using a flatness preparation process. Multidimensional flatness parameters of the multiple rock samples were collected, and various mechanical properties, water content and density of the multiple rock samples were measured to generate a multidimensional dataset. Based on the smoothness parameters, lithology, water content, and density, a multidimensional input feature vector is constructed; An association model is constructed by integrating a random forest model and a backpropagation neural network. The multidimensional dataset is used as training data to train the association model. The model hyperparameters are optimized by cross-validation to dynamically adjust the feature weights for lithology. The association model uses the multidimensional input feature vector as input to predict the mechanical indices corresponding to the input features. A differential correction model is constructed to correct the output of the correlation model for different lithologies. Based on the differential correction model, the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested is obtained.

2. The method according to claim 1, characterized in that, The association model is specifically used for: The interaction effects between different parameters in the multidimensional input feature vector are identified using the Multivariate Adaptive Regression Spline (MARS) algorithm.

3. The method according to claim 1, characterized in that, The dynamic adjustment of feature weights for lithology includes: Weighting factors for water content and density were set for lithology. Specifically, the weight of water content was increased when the rock sample was soft rock, and the weight of density was increased when the rock sample was hard rock. In the random forest model, weight thresholds for water content and density are set.

4. The method according to claim 1, characterized in that, The process of modifying the output of the correlation model using the differentiated correction model for different lithologies includes: The multidimensional input feature vector is stratified, and the stratified data is then standardized. For soft rock, a first correction formula is constructed with uniaxial compressive strength as the dependent variable and standardized roughness parameters, water content and density as independent variables. The first correction formula contains multiple lithological correlation coefficients to be solved. Ridge regression is applied to the first modified formula to construct a loss function that includes an L2 regularization term; The regularization parameter in the loss function is optimized by 5-fold cross-validation to obtain multiple lithological correlation coefficients obtained from the fitting.

5. The method according to claim 4, characterized in that, After standardizing the stratified data, the process further includes: For hard rock, multiple target features sensitive to hard rock are screened from the standardized data, and a second correction formula is constructed with uniaxial compressive strength as the dependent variable and the multiple target features as independent variables. The second correction formula includes multiple lithological correlation coefficients to be solved. The second modified formula is fitted by least squares method, and a three-dimensional particle flow model PFC is constructed. Discrete element simulation is performed through the three-dimensional particle flow model PFC to calibrate the lithological correlation coefficient corresponding to density. The correlation model is corrected by expanding the training dataset based on discrete element simulation.

6. The method according to claim 1, characterized in that, The process of obtaining the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested includes: The multidimensional input feature vector of the rock sample to be tested is received through a visualization query platform; The trained prediction model is invoked to output the predicted mechanical index value corresponding to the currently received multi-dimensional input feature vector, wherein the prediction model includes the correlation model and the differential correction model; Within a preset flatness range, the predicted mechanical index values ​​are used to generate equally spaced data points in unit step sizes, and the predicted mechanical index curve is obtained by difference fitting.

7. The method according to claim 6, characterized in that, After obtaining the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested, the method further includes: The mechanical index prediction curve is displayed through the interactive interface of the visualization query platform; Within the preset flatness range, a strength correction coefficient table is generated.

8. The method according to claim 1, characterized in that, The preparation of multiple rock samples with different end-face flatness using a flatness preparation process includes: By combining high-precision grinding and polishing equipment with pressure sensors, the flatness of the end face of different rock samples can be adjusted.

9. A system for determining the correlation between the end-face flatness and mechanical properties of a rock sample, characterized in that, Includes the following modules: The acquisition module is used to prepare multiple rock samples with different end face flatness using a flatness preparation process, acquire multidimensional flatness parameters of the multiple rock samples, and measure various mechanical properties, water content and density of the multiple rock samples to generate a multidimensional dataset. A construction module is used to construct a multidimensional input feature vector based on the smoothness parameters, lithology, water content, and density; The prediction module is used to construct an association model by integrating a random forest model and a backpropagation neural network, train the association model using the multidimensional dataset as training data, and optimize the model hyperparameters through cross-validation to dynamically adjust the feature weights for lithology. The association model uses the multidimensional input feature vector as input to predict the mechanical indicators corresponding to the input features. The correction module is used to construct a differential correction model, which corrects the output of the correlation model for different lithologies, and obtains the mechanical index prediction curve corresponding to the multidimensional input feature vector of the rock sample to be tested based on the differential correction model.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for determining the correlation between the flatness of the rock sample end face and mechanical properties as described in any one of claims 1-8.

Images