Feature analysis-machine learning based discrete element contact parameter prediction method and system

Abstract

The application discloses a discrete element contact parameter prediction method based on feature analysis-machine learning, comprising the following steps: selecting a discrete element linear parallel bond model for contact parameter calibration according to soft rock particle characteristics and mold material parameters, and determining model parameters and value ranges; obtaining contact parameters of the discrete element linear parallel bond model based on a small amount of uniaxial compression and Brazilian splitting tests, and constituting a contact parameter-soft rock macro feature low-fidelity dataset based on the contact parameters; determining key contact parameters affecting the soft rock macro feature based on grey relational analysis (GRA), Pearson correlation analysis and mutual information (MI); obtaining a contact parameter-macro feature high-fidelity dataset based on a large amount of uniaxial compression and Brazilian splitting tests; establishing a key contact parameter prediction model of the discrete element linear parallel bond model based on machine learning; and determining an optimal prediction model of the key contact parameters of the linear parallel bond model. Corresponding systems, electronic devices and computer readable storage media are also disclosed.

Description

Technical Field

[0001] This invention relates to the fields of rock mechanics and materials science testing technology, and in particular to a discrete element contact parameter prediction method and system based on feature analysis-machine learning. Background Technology

[0002] The main problems in current discrete element contact parameter prediction are as follows:

[0003] (1) Low efficiency, long time consumption, and large computational load: Traditional discrete element contact parameter calibration methods usually require a large number of experiments and numerical simulations, resulting in low efficiency and long time consumption. Especially in complex rock mechanics problems, the calibration process of contact parameters often requires repeated adjustments and verifications, resulting in a huge amount of computation.

[0004] (2) Single-level performance index makes it difficult to select the optimal model: In the traditional parameter calibration process, the accuracy of the model is usually evaluated by a single mechanical performance index, which may lead to poor performance of the model in other performance indexes, making it difficult to select the optimal model.

[0005] (3) Lack of efficient parameter prediction methods: Most existing discrete element contact parameter prediction methods rely on empirical formulas or manual adjustments, lacking systematicity and automation, and are difficult to cope with complex rock mechanics problems. Summary of the Invention

[0006] In view of this, to address the problems in the background technology, this invention proposes a discrete element contact parameter prediction method and system based on feature analysis and machine learning. By combining methods such as grey relational analysis (GRA), Pearson correlation analysis, and mutual information (MI), key contact parameters affecting the macroscopic mechanical properties of soft rock are determined, and a contact parameter prediction model is established using machine learning algorithms (such as BPNN, SVR, XGBoost, etc.). Furthermore, a particle swarm optimization (PSO) algorithm is introduced to optimize the parameters of the machine learning model, thereby improving the model's prediction accuracy and generalization ability. This method and system exhibit good prediction accuracy and generalization ability, providing a new approach to contact parameter prediction. The method mainly includes the following steps: First, low-fidelity datasets are obtained through uniaxial compression tests and Brazilian splitting tests to determine the main microscopic parameters affecting the macroscopic characteristics (strength and deformation) of red bed soft rock. Second, based on grey relational analysis (GRA), Pearson correlation analysis, and mutual information (MI), key contact parameters affecting the macroscopic characteristics of soft rock are determined, and a high-fidelity dataset is obtained. Finally, a key contact parameter prediction model based on a linear parallel bond model is established, and the optimal prediction model is determined based on prediction accuracy evaluation.

[0007] The first aspect of this invention is to provide a discrete element contact parameter prediction method based on feature analysis-machine learning, which predicts discrete element contact parameters based on an optimal prediction model for key contact parameters of a linear parallel bonding model. The method includes:

[0008] S1. Based on the characteristics of soft rock particles and the material parameters of the mold, select the discrete element linear parallel bond model for contact parameter calibration, and determine the model parameters and the range of model parameter values ​​for the discrete element linear parallel bond model to be calibrated.

[0009] S2. Contact parameters of the discrete element linear parallel bond model are obtained based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parameters affecting the macroscopic characteristics of soft rock. The main micro-parameters are used to construct a low-fidelity dataset of contact parameters and macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation.

[0010] S3. Based on grey relational analysis (GRA), Pearson correlation analysis, and mutual information (MI), the key contact parameters affecting the macroscopic characteristics of soft rock were determined.

[0011] S4, a high-fidelity dataset of contact parameters and macroscopic features obtained from a large number of uniaxial compression and Brazilian splitting experiments;

[0012] S5, a key contact parameter prediction model for discrete element linear parallel bonding model based on machine learning;

[0013] S6, determine the optimal prediction model for key contact parameters of the linear parallel bond model.

[0014] Preferably, S2 includes:

[0015] S21, Obtain the first contact parameter - soft rock macroscopic features low-fidelity dataset corresponding to the discrete element linear parallel bonding model based on uniaxial compression test numerical simulation;

[0016] S22, Obtain the low-fidelity dataset of the second contact parameters-macroscopic features corresponding to the discrete element linear parallel bonding model based on the Brazilian splitting test numerical simulation; among which, the tensile strength σ is selected. t As an indicator of the Brazilian splitting test;

[0017] S23, the first contact parameter-macro feature low-fidelity dataset and the second contact parameter-macro feature low-fidelity dataset are concatenated to form the contact parameter-soft rock macro feature low-fidelity dataset.

[0018] Preferably, S21 includes:

[0019] (1) Determine the parameters required for the uniaxial compression test, including:

[0020] Geometric parameters include: axial length L of the specimen, width W of the specimen, porosity n, radius R of the specimen, length-to-radius ratio L / R of the specimen, and maximum size R of the particle or pore feature unit. max and the minimum size R of the particle or pore characteristic unit max The ratio of the maximum to minimum size of a particle or pore feature unit, Rmax / Rmin;

[0021] Material and deformation parameters, including: particle density ρb, particle-to-particle linear effective modulus E b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio and particle-particle friction coefficient μ;

[0022] Strength parameters, including: parallel bond tensile strength σ t Parallel bond shear strength τ c and internal friction angle ;

[0023] The coefficient of friction μ between particles and the wall bw ;

[0024] (2) Determine the test factor level and set the uniaxial compression test parameters under different test factor levels;

[0025] Determine the particle-particle linear effective modulus E b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio Particle-to-particle friction coefficient μ, parallel bond tensile strength σ t Parallel bond shear strength τ c and internal friction angle The upper and lower limits of the variable parameters used in uniaxial compression numerical experiments; and the different values ​​at four experimental factor levels;

[0026] (3) A uniaxial compression test is conducted based on the orthogonal test rules, and a numerical simulation based on the uniaxial compression test is performed to obtain the first contact parameter-soft rock macroscopic characteristics low-fidelity data corresponding to the discrete element linear parallel bonding model based on the uniaxial compression test; wherein, the uniaxial compression test includes: generating particles in a container with the geometric parameters formed by the sample, wherein the particle size satisfies a Gaussian normal distribution, and setting the particle-particle and particle-wall contacts as linear model contacts to simulate the occurrence conditions of the rock; the numerical simulation based on the uniaxial compression test includes: modifying the contact parameters in the linear model according to the test parameters, and setting the particle-particle contact as a linear parallel bonding model contact; loading the upper and lower loading plates of the container at a constant speed v z Numerical simulations based on uniaxial compression tests were conducted under pressure; triaxial tests under low confining pressure were used as an approximation of uniaxial tests.

[0027] Preferably, S3 includes:

[0028] S31, Grey Relational Analysis (GRA) is used to determine the microscopic parameters that have the most influence on the macroscopic characteristics of soft rock. Based on Grey Relational Analysis (GRA), Pearson correlation analysis, and mutual information (MI), key contact parameters are initially determined from the most influential microscopic parameters, including:

[0029] (1) The micro-parameters are quantified using correlation metrics through grey relational analysis (GRA), including:

[0030] A. Determine the analytical sequence; the determined analytical sequence includes:

[0031] Each macroscopic characteristic value of each uniaxial compression numerical test sample is used as the reference sequence for that uniaxial compression numerical test sample.

[0032] The contact parameter values ​​of the discrete element linear parallel bond model of each uniaxial compression numerical test sample are used as the comparison sequence of that uniaxial compression numerical test sample.

[0033] B. Calculate and analyze the correlation coefficients and correlation degrees of various detailed parameters in the sequence; including:

[0034] Based on the reference and comparison sequences of all uniaxial compression numerical test samples, the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for each uniaxial compression numerical test sample are calculated.

[0035] Based on the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for all uniaxial compression numerical test samples, the correlation degree between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model is calculated.

[0036] C. Compressive strength Elastic modulus Poisson's ratio Ranking of influencing factors and determining the most influential micro-parameters;

[0037] (2) Based on Pearson correlation analysis and mutual information (MI), the correlation between the most influential microscopic parameters and macroscopic features was analyzed, and the correlation between the most influential microscopic parameters and macroscopic features of soft rock was obtained by comparing the results of grey relational analysis (GRA). Among them, Pearson is used to measure the degree of linear correlation between two variables. The calculation formula of Pearson correlation analysis is as shown in equation (6), and the calculation formula of mutual information (MI) is as shown in equation (7).

[0038] (6)

[0039] In equation (6), For the discrete element linear parallel bonding model, the first The contact parameter and the first Pearson correlation coefficients of macroscopic features for In the first uniaxial compression numerical test sample, the discrete element linear parallel bond model of the first... The average value of each contact parameter for In the uniaxial compression numerical test sample, the first The average of a macroscopic characteristic;

[0040] (7)

[0041] In equation (7), For the discrete element linear parallel bonding model, the first The contact parameter and the first The mutual information value of a macroscopic feature, for and The joint probability mass function, For the discrete element linear parallel bonding model, the first Macroeconomic characteristics All possible values, For the discrete element linear parallel bonding model, the first Contact parameters All possible values, The first discrete-element linear parallel bond model represents the... A macroscopic characteristic, The first discrete-element linear parallel bond model represents the... One contact parameter, The first discrete-element linear parallel bond model represents the... Contact parameters Values The probability, i.e. marginal probability mass function The first discrete-element linear parallel bond model represents the... Macroeconomic characteristics Values The probability, i.e. The marginal probability mass function;

[0042] (3) Based on grey relational analysis (GRA), Pearson analysis, and mutual information analysis (MI), correlation analysis was performed on the numerical test results of uniaxial compression.

[0043] S32, based on impact analysis, the key contact parameters affecting the macroscopic characteristics of soft rock were preliminarily determined, thereby identifying the main mesoscopic parameters affecting the elastic modulus E and Poisson's ratio v. and Both show a linear relationship; affecting compressive strength σ f and tensile strength σ t The main microscopic parameter is the parallel bond shear strength τ. c Parallel bond shear strength τ c It can be determined by the tension-compression ratio k.

[0044] Preferably, S4 includes:

[0045] S41, the first high-fidelity dataset was obtained based on a large number of uniaxial compression numerical experiments;

[0046] S42, based on a large number of Brazilian splitting experiments to obtain a second high-fidelity dataset;

[0047] S43, obtain a high-fidelity dataset of contact parameters and macroscopic features based on the first high-fidelity dataset and the second high-fidelity dataset.

[0048] Preferably, S5 includes:

[0049] S51, based on elastic modulus Compressive strength Tensile strength σ t The tensile-compression ratio k is used to construct key contact parameters, including the effective modulus of particle-particle bonding. Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c The dataset;

[0050] S52, the high-fidelity dataset of contact parameters-macroscopic features is divided into training set T according to a certain ratio. train and test set T testThe training set T train Used to build machine learning models and test sets T test Used to test the performance of machine learning models;

[0051] S53, establish a system based on training set T train Machine learning models, including:

[0052] (1) Predicting key contact parameter particle-particle bonding effective modulus based on typical machine learning prediction algorithms Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c A machine learning model is established; wherein, the typical machine learning prediction algorithms include backpropagation neural network, support vector regression machine and XGBoost;

[0053] (2) The hyperparameters of the discrete element linear parallel bonding model are optimized based on the particle swarm optimization algorithm to obtain the optimal hyperparameters;

[0054] (3) The training set T train Input the machine learning model and train the machine learning model based on the optimal hyperparameters.

[0055] Preferably, S6 includes:

[0056] S61, calculate the prediction accuracy and error of the machine learning model;

[0057] Calculate the goodness of fit R 2 Mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE) are used as the prediction accuracy and error metrics of the machine learning model; where:

[0058] The goodness of fit R 2 The formula is shown in equation (9) below:

[0059] (9)

[0060] The mean square error (MSE) is expressed in the following formula (10):

[0061] (10)

[0062] The root mean square error (RMSE) formula is shown in equation (11):

[0063] (11)

[0064] The formula for the mean absolute error (MAE) is shown in equation (12):

[0065] (12)

[0066] In equations (9)-(12), For the sample size, for , and τ c The measured value, for and τ c The predicted value, for , and τ c The average of the measured values;

[0067] S62, evaluate the quality of the machine learning model based on the prediction accuracy and error hierarchy, thereby determining the optimal prediction model for the key contact parameters of the linear parallel bonding model; wherein the characteristics of each index are: R 2 The range is between 0 and 1, and the closer the value is to 1, the higher the precision; MSE, RMSE and MAE are all greater than 0, and the closer the value is to 0, the smaller the error.

[0068] A second aspect of the present invention provides a vibration suppression system for an unmanned aerial vehicle (UAV) used for track interlayer inspection, for implementing the method of the first aspect, the system comprising:

[0069] The model and model parameter determination module (101) is used to select the discrete element linear parallel bond model for contact parameter calibration based on the characteristics of soft rock particles and the material parameters of the mold, and to determine the model parameters and the range of model parameter values ​​of the discrete element linear parallel bond model to be calibrated.

[0070] The contact parameter test acquisition module (102) is used to acquire the contact parameters of the discrete element linear parallel bond model based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parallel parameters that affect the macroscopic characteristics of soft rock, and the main micro-parallel parameters are used to form a low-fidelity dataset of contact parameters-macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation.

[0071] The key contact parameter determination module (103) is used to determine the key contact parameters affecting the macroscopic characteristics of soft rock based on grey relational analysis (GRA), Pearson correlation analysis and mutual information (MI);

[0072] A high-fidelity dataset acquisition module (104) is used to acquire a high-fidelity dataset of contact parameters and macroscopic features based on a large number of uniaxial compression and Brazilian splitting experiments.

[0073] The machine learning model building module (105) is used to build a key contact parameter prediction model for a discrete element linear parallel bonding model based on machine learning.

[0074] The optimal prediction model determination module (106) is used to determine the optimal prediction model for key contact parameters of the linear parallel bonding model.

[0075] A third aspect of the present invention provides an electronic device including a processor and a memory, the memory storing a plurality of instructions, the processor being configured to read the instructions and execute the method as described in the first aspect.

[0076] A fourth aspect of the present invention provides a computer-readable storage medium storing a plurality of instructions which can be read by a processor and executed as described in the first aspect.

[0077] The beneficial effects of the method and system of the present invention are as follows:

[0078] This method addresses the low efficiency of contact parameter prediction in linear parallel bond models and the difficulty in selecting the optimal ML model based on a single-level performance index. First, a small number of uniaxial compression and Brazilian splitting tests are conducted to obtain a low-fidelity dataset of contact parameters and rock macroscopic characteristics from the linear parallel bond model. Second, based on GRA, Pearson, and MI, key contact parameters affecting rock macroscopic characteristics are identified. Finally, a large number of experiments are conducted to obtain a high-fidelity dataset, a prediction ML model is established, and the optimal ML model is determined based on the accuracy evaluation of the ML model. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of the present invention, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0080] Figure 1 This is a flowchart of a discrete element contact parameter prediction method based on feature analysis-machine learning provided according to an embodiment of the present invention;

[0081] Figure 2 This is a system architecture diagram of a discrete element contact parameter prediction system based on feature analysis-machine learning provided according to an embodiment of the present invention;

[0082] Figure 3 (a)-(c) are chord diagrams showing the gamma correlation between contact parameters and macroscopic properties under three different algorithms;

[0083] Figure 4 This is a trend analysis diagram of the influence of three types of correlation analysis algorithms and microscopic parameters on the macroscopic characteristics of soft rock particles;

[0084] Figure 5 (a)-(h) are dotted-line graphs showing the trend of the influence of micro-parameters on macro-features;

[0085] Figure 6 The relationship between each combination of E, σf, σt, and k in the dataset and the key contact parameters;

[0086] Figure 7 (a)-(d) represent the fitting results of the ML model's predicted values ​​and measured values ​​on the training set;

[0087] Figure 8 (a)-(d) are scatter plots of the predictions of the ML model on the test set;

[0088] Figure 9 (a)-(d) show the prediction accuracy and error evaluation results of the MAE of each ML model on the test set;

[0089] Figure 10 (a)-(d) show the prediction accuracy and error evaluation results of each ML model's RMSE on the test set;

[0090] Figure 11 (a)-(d) show the prediction accuracy and error evaluation results of each ML model's MSE on the test set;

[0091] Figure 12 (a)-(d) show the prediction accuracy and error evaluation results of each ML model on the test set using R2;

[0092] Figure 13 This is a structural diagram of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0093] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0094] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0095] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0096] Example 1

[0097] like Figure 1 As shown, this embodiment provides a discrete element contact parameter prediction method based on feature analysis-machine learning. It predicts discrete element contact parameters based on the optimal prediction model for key contact parameters of a linear parallel bonding model. The method includes:

[0098] S1. Based on the characteristics of soft rock particles and the material parameters of the mold, select the discrete element linear parallel bond model for contact parameter calibration, and determine the model parameters and the range of model parameter values ​​for the discrete element linear parallel bond model to be calibrated.

[0099] S2. Contact parameters of the discrete element linear parallel bond model are obtained based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parameters affecting the macroscopic characteristics of soft rock. The main micro-parameters are used to construct a low-fidelity dataset of contact parameters and macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation.

[0100] In a preferred embodiment, S2 includes:

[0101] S21, Obtain the low-fidelity dataset of the first contact parameters—macroscopic characteristics of soft rock—corresponding to the discrete element linear parallel bond model based on uniaxial compression tests and numerical simulation; including:

[0102] (1) Determine the parameters required for the uniaxial compression test, including:

[0103] Geometric parameters include: axial length L of the specimen, width W of the specimen, porosity n (a parameter describing the amount of pores inside the material, reflecting the microstructure characteristics of the material, and having a significant impact on the mechanical properties of the material), radius R of the specimen, length-to-radius ratio L / R of the specimen (a parameter characterizing the aspect ratio of the specimen; different aspect ratios may result in different mechanical responses of the material in uniaxial compression tests), and maximum size R of the particle or pore characteristic unit. max and the minimum size R of the particle or pore characteristic unit maxThe ratio of the maximum to minimum size of a particle or pore characteristic unit, Rmax / Rmin (used to reflect the non-uniformity of the internal structural dimensions of a material; it is a microstructural parameter. The larger this ratio, the greater the difference in the internal structural dimensions of the material, which may affect the material's strength, deformation, and other macroscopic mechanical properties).

[0104] Material and deformation parameters, including: particle density ρb, particle-to-particle linear effective modulus E b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio and particle-particle friction coefficient μ;

[0105] Strength parameters, including: parallel bond tensile strength σ t Parallel bond shear strength τ c and internal friction angle ;

[0106] The coefficient of friction μ between particles and the wall bw In uniaxial compression numerical tests, there is contact between particles and the wall, requiring μ to be... bw Also take μ into account bw Set it to 0.1.

[0107] (2) Determine the test factor level and set the uniaxial compression test parameters under different test factor levels;

[0108] The coefficient of friction μ between particles and the wall bw Set to 0.1;

[0109] The levels of the four experimental factors were determined to be 1 to 4.

[0110] Determine the particle-particle linear effective modulus E b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio Particle-to-particle friction coefficient μ, parallel bond tensile strength σ t Parallel bond shear strength τ c and internal friction angle The upper and lower limits of the variable parameters used in uniaxial compression numerical experiments; and the different values ​​at four experimental factor levels;

[0111] (3) A uniaxial compression test is conducted based on the orthogonal test rules, and a numerical simulation based on the uniaxial compression test is performed to obtain the first contact parameter-soft rock macroscopic characteristics low-fidelity data corresponding to the discrete element linear parallel bonding model based on the uniaxial compression test; wherein, the uniaxial compression test includes: generating particles in a container with the geometric parameters formed by the sample, wherein the particle size satisfies a Gaussian normal distribution, and setting the particle-particle and particle-wall contacts as linear model contacts to simulate the occurrence conditions of the rock; the numerical simulation based on the uniaxial compression test includes: modifying the contact parameters in the linear model according to the test parameters, and setting the particle-particle contact as a linear parallel bonding model contact; loading the upper and lower loading plates of the container at a constant speed v z Numerical simulations based on uniaxial compression tests were conducted under applied pressure. In this embodiment, a triaxial test under low confining pressure was used to approximate the uniaxial test for easier testing of the Poisson's ratio v.

[0112] elastic modulus The calculation formula is shown in equation (1) below:

[0113] (1)

[0114] In equation (1), σ represents stress, i.e., the force per unit area; It represents strain, that is, the relative amount of material deformation;

[0115] The formula for calculating Poisson's ratio v is shown in equation (2):

[0116] (2)

[0117] In equation (2), s It represents transverse strain, which is the deformation of a material in the direction perpendicular to the external force. This represents axial strain, which is the deformation of a material in the direction of an external force.

[0118] S22, Obtain the low-fidelity dataset of the second contact parameters-macroscopic features corresponding to the discrete element linear parallel bonding model based on the Brazilian splitting test numerical simulation; among which, the tensile strength σ is selected. t As an indicator of the Brazilian splitting test.

[0119] The formula for calculating tensile strength is shown in equation (3):

[0120] (3)

[0121] In equation (3), F2 represents the applied tensile force, and A2 represents the tensile area.

[0122] S23, the first contact parameter-macro feature low-fidelity dataset and the second contact parameter-macro feature low-fidelity dataset are concatenated to form the contact parameter-soft rock macro feature low-fidelity dataset.

[0123] S3. Based on grey relational analysis (GRA), Pearson correlation analysis and mutual information (MI), the key contact parameters affecting the macroscopic characteristics of soft rock were determined.

[0124] In a preferred embodiment, S3 includes:

[0125] S31, Grey Relational Analysis (GRA) is used to determine the microscopic parameters that have the greatest influence on the macroscopic characteristics of soft rock. Based on Grey Relational Analysis (GRA), Pearson correlation analysis, and mutual information (MI), key contact parameters are initially determined from the most influential microscopic parameters, including:

[0126] (1) The micro-parameters are quantified using correlation metrics through grey relational analysis (GRA), including:

[0127] A. Determine the sequence to be analyzed;

[0128] In this embodiment, the determined analysis sequence includes:

[0129] Each macroscopic characteristic value of each uniaxial compression numerical test sample is used as the reference sequence for that uniaxial compression numerical test sample.

[0130] The contact parameter values ​​of the discrete element linear parallel bond model of each uniaxial compression numerical test sample are used as the comparison sequence of that uniaxial compression numerical test sample.

[0131] B. Calculate and analyze the correlation coefficients and correlation degrees of various detailed parameters in the sequence; including:

[0132] Based on the reference and comparison sequences of all uniaxial compression numerical test samples, the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for each uniaxial compression numerical test sample are calculated.

[0133] Based on the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for all uniaxial compression numerical test samples, the correlation degree between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model is calculated.

[0134] In this embodiment, it is assumed that... The numerical test of uniaxial compression yielded the following results. The uniaxial compression numerical test sample, for the first The uniaxial compression numerical test sample, the first The reference sequence for each uniaxial compression numerical test sample is as follows: , They represent the first Compressive strength of a single uniaxial compression numerical test specimen Elastic modulus Poisson's ratio ;

[0135] No. The comparison sequence of the uniaxial compression numerical test samples is as follows: , This represents the total number of contact parameters in the discrete-element linear parallel bond model. This is an index of the contact parameters in the discrete-element linear parallel bond model. For the first The discrete element linear parallel bond model of the first uniaxial compression numerical test sample One contact parameter value;

[0136] Calculate the first In the uniaxial compression numerical test samples, the discrete element linear parallel bond model of the first... The contact parameter and the first The correlation coefficient of the macroscopic features is given by the formula shown in equation (4):

[0137] (4);

[0138] In equation (4), For the discrete element linear parallel bonding model, the first The contact parameter and the first The correlation coefficient of a macroscopic feature In order to be in In the uniaxial compression numerical test sample, the first The minimum absolute difference between a macroscopic feature and all contact parameter values. In order to be in In the uniaxial compression numerical test samples, the discrete element linear parallel bond model of the first... The maximum absolute difference between a macroscopic feature and all contact parameter values. The resolution coefficient, For the first The discrete element linear parallel bond model of the first uniaxial compression numerical test sample One macroscopic characteristic value, For the first The discrete element linear parallel bond model of the first uniaxial compression numerical test sample Contact parameter values.

[0139] Based on the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for all uniaxial compression numerical test samples, the correlation degree between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model is calculated, and the formula is shown in Equation (5):

[0140] (5);

[0141] In equation (5), For the discrete element linear parallel bonding model, the first The contact parameter and the first The degree of correlation of macro-characteristics.

[0142] C. Compressive strength Elastic modulus Poisson's ratio Ranking of influencing factors and determining the most influential micro-parameters;

[0143] In this embodiment, Grey Relational Analysis (GRA) is a multi-factor statistical analysis method used to determine the degree of correlation between factors. The following is its basic formula:

[0144] a. Data preprocessing (standardization or normalization)

[0145] The raw data is usually averaged or initialized to eliminate differences in dimensions.

[0146] Mean-orientation: ;

[0147] Initialization (based on the first data point): ;

[0148] b. Calculate the correlation coefficient

[0149] Let the reference sequence be The comparison sequence is The correlation coefficient formula is: ;

[0150] in: The absolute difference between the reference sequence and the comparison sequence at point k; : Global minimum difference; : Global maximum difference; The resolution coefficient (usually taken as 0.5, in the range (0,1)).

[0151] c. Calculate the correlation degree

[0152] The degree of correlation is the average of the correlation coefficients, reflecting the overall level of correlation.

[0153] ;

[0154] d. Relevance Ranking

[0155] By correlation Sort from largest to smallest; the larger the value, the stronger the correlation between the comparison sequence and the reference sequence.

[0156] (2) Based on Pearson correlation analysis and mutual information (MI), the correlation between the most influential microscopic parameters and macroscopic characteristics is analyzed. Compared with the results of GRA analysis, the correlation between the most influential microscopic parameters and macroscopic characteristics of soft rock is obtained more accurately. Among them, Pearson is used to measure the degree of linear correlation between two variables. The calculation formula of Pearson correlation analysis is as shown in Equation (6), and the calculation formula of mutual information (MI) is as shown in Equation (7).

[0157] The calculation process of Pearson correlation analysis is shown in equation (6):

[0158] (6)

[0159] In equation (6), For the discrete element linear parallel bonding model, the first The contact parameter and the first Pearson correlation coefficients of macroscopic features for In the first uniaxial compression numerical test sample, the discrete element linear parallel bond model of the first... The average value of each contact parameter for In the uniaxial compression numerical test sample, the first The average value of a macroscopic feature.

[0160] The formula for calculating mutual information (MI) is shown in equation (7):

[0161] (7);

[0162] In equation (7), For the discrete element linear parallel bonding model, the first The contact parameter and the first The mutual information value of a macroscopic feature, for and The joint probability mass function, For the discrete element linear parallel bonding model, the first Macroeconomic characteristics All possible values, For the discrete element linear parallel bonding model, the first Contact parameters All possible values, The first discrete-element linear parallel bond model represents the... A macroscopic characteristic, The first discrete-element linear parallel bond model represents the... One contact parameter, The first discrete-element linear parallel bond model represents the... Contact parameters Values The probability, i.e. marginal probability mass function The first discrete-element linear parallel bond model represents the... Macroeconomic characteristics Values The probability, i.e. The marginal probability mass function.

[0163] Mutual information (MI) is a metric in information theory that measures the statistical dependence between two random variables, reflecting the amount of information one variable contains about the other. The following is its core formula and explanation:

[0164] Mutual information definition

[0165] The formula for mutual information I(X;Y) is: ;

[0166] In the formula: p(x,y) represents the joint probability distribution of X and Y; p(x) and p(y) represent the marginal probability distributions of X and Y; the logarithmic base is usually 2 (in bits) or the natural logarithm e (in knights).

[0167] (3) Correlation analysis of uniaxial compression numerical test results based on GRA, Pearson and MI.

[0168] S32, based on impact analysis, the key contact parameters affecting the macroscopic characteristics of soft rock were preliminarily determined, thereby identifying the main mesoscopic parameters affecting the elastic modulus E and Poisson's ratio v. and Both show a linear relationship; while the compressive strength σ f、 Tensile strength σ t The main microscopic parameter is the parallel bond shear strength τ. c Parallel bond shear strength τ c It can then be determined by the tension-compression ratio k.

[0169] compressive strength σ f The calculation formula is shown in equation (8):

[0170] (8);

[0171] In equation (8), F1 represents the applied force and A1 represents the area under pressure.

[0172] In this embodiment, S32 includes: analyzing the influence trend of macroscopic features of soft rock particles based on three types of correlation analysis algorithms and microscopic parameters, obtaining the main microscopic parameters affecting macroscopic features, and further determining the key contact parameters affecting macroscopic features.

[0173] S4, a high-fidelity dataset of contact parameters and macroscopic features obtained from a large number of uniaxial compression and Brazilian splitting experiments;

[0174] In a preferred embodiment, S4 includes:

[0175] S41, the first high-fidelity dataset was obtained based on a large number of uniaxial compression numerical experiments;

[0176] S42, based on a large number of Brazilian splitting experiments to obtain a second high-fidelity dataset;

[0177] S43, obtain a high-fidelity dataset of contact parameters and macroscopic features based on the first high-fidelity dataset and the second high-fidelity dataset.

[0178] S5, a key contact parameter prediction model for discrete element linear parallel bonding model based on machine learning;

[0179] In a preferred embodiment, S5 includes:

[0180] S51, based on elastic modulus Compressive strength Tensile strength σ t The tensile-compression ratio k is used to construct key contact parameters, including the effective modulus of particle-particle bonding. Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c The dataset;

[0181] In this embodiment, by analyzing the relationship between different macroscopic characteristics of soft rock and key contact parameters, the effective modulus of particle-particle bonding, a key contact parameter, is established. Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c The dataset D = {(x, y)}, where x is the input feature of the prediction model, and is composed of the elastic modulus. Compressive strength Tensile strength σ t The output characteristics are composed of the tensile-compression ratio k, and the key contact parameter is the effective modulus of particle-particle bonding. Particle-to-particle bond stiffness ratio and parallel bond shear strength τc composition;

[0182] S52, the high-fidelity dataset of contact parameters-macroscopic features is divided into training set T according to a certain ratio. train and test set T test The training set T train Used to build machine learning models and test sets T test Used to test the performance of machine learning models.

[0183] In this embodiment, the contact parameter-macro feature high-fidelity dataset is divided into a training set and a test set in a 7:3 ratio, wherein the training set is used to build a machine learning model and the test set is used to test the performance of the machine learning model.

[0184] S53, establish a system based on training set T train Machine learning models, including:

[0185] (1) Predicting key contact parameter particle-particle bonding effective modulus based on typical machine learning prediction algorithms Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c Establish machine learning models; wherein, the typical machine learning prediction algorithms include Back Propagation Neural Network (BPNN), Support Vector Machine (SVM), and XGBoost;

[0186] (2) The hyperparameters of the discrete element linear parallel bonding model are optimized based on the Particle Swarm Optimization (PSO) algorithm to obtain the optimal hyperparameters;

[0187] (3) The training set T train Input the machine learning model and train the machine learning model based on the optimal hyperparameters.

[0188] S6, determine the optimal prediction model for key contact parameters of the linear parallel bond model.

[0189] Since the prediction results on the training set only represent the predictive ability of the machine learning model during the modeling process, and cannot reflect the generalization ability of the model, the predictive performance of the machine learning model should be judged through the test set.

[0190] In a preferred embodiment, S6 includes:

[0191] S61, calculate the prediction accuracy and error of the machine learning model;

[0192] Calculate the goodness of fit R 2 Mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE) are used as the prediction accuracy and error metrics of the machine learning model; where:

[0193] The goodness of fit R 2 The formula is shown in equation (9) below:

[0194] (9)

[0195] The mean square error (MSE) is expressed in the following formula (10):

[0196] (10)

[0197] The root mean square error (RMSE) formula is shown in equation (11):

[0198] (11)

[0199] The formula for the mean absolute error (MAE) is shown in equation (12):

[0200] (12)

[0201] In equations (9)-(12), For the sample size, for , and τ c The measured value, for and τ c The predicted value, for , and τ c The average of the measured values;

[0202] S62, evaluate the quality of the machine learning model based on the prediction accuracy and error hierarchy, thereby determining the optimal prediction model for the key contact parameters of the linear parallel bonding model; wherein each indicator characteristic is R 2 The range is between 0 and 1, and the closer the value is to 1, the higher the precision; MSE, RMSE and MAE are all greater than 0, and the closer the value is to 0, the smaller the error.

[0203] Example 2

[0204] like Figure 2 As shown, this embodiment provides a vibration suppression system for an unmanned aerial vehicle (UAV) used for track interlayer inspection, for implementing the method of Embodiment 1. The system includes:

[0205] The model and model parameter determination module 101 is used to select a discrete element linear parallel bond model for contact parameter calibration based on the characteristics of soft rock particles and the material parameters of the mold, and to determine the model parameters and the range of model parameter values ​​for the discrete element linear parallel bond model to be calibrated.

[0206] The contact parameter test acquisition module 102 is used to acquire the contact parameters of the discrete element linear parallel bond model based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parallel parameters that affect the macroscopic characteristics of soft rock, and the main micro-parallel parameters are used to construct a low-fidelity dataset of contact parameters-macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation.

[0207] The key contact parameter determination module 103 is used to determine the key contact parameters affecting the macroscopic characteristics of soft rock based on grey relational analysis (GRA), Pearson correlation analysis and mutual information (MI);

[0208] High-fidelity dataset acquisition module 104 is used to acquire a high-fidelity dataset of contact parameters and macroscopic features based on a large number of uniaxial compression and Brazilian splitting experiments.

[0209] The machine learning model building module 105 is used to build a key contact parameter prediction model for a discrete element linear parallel bonding model based on machine learning.

[0210] The optimal prediction model determination module 106 is used to determine the optimal prediction model for key contact parameters of the linear parallel bonding model.

[0211] Application Examples

[0212] This embodiment discloses a discrete element method for predicting contact parameters based on machine learning, comprising the following steps: First, low-fidelity datasets are obtained through uniaxial compression tests and Brazilian splitting tests to determine the main microscopic parameters affecting the macroscopic characteristics (strength and deformation) of red bed soft rock. Second, based on GRA, Pearson, and MI, key contact parameters affecting the macroscopic characteristics of soft rock are determined, and high-fidelity datasets are obtained. Finally, a linear parallel bond model is established to predict key contact parameters, and the optimal prediction model is determined based on prediction accuracy evaluation.

[0213] The embodiments and implementation process of the complete method according to the present invention are as follows:

[0214] 1) Select discrete element linear parallel bonding based on the characteristics of soft rock particles and the material parameters of the mold. Determine the parameters and value range of the discrete element contact model to be calibrated.

[0215] 2) Uniaxial compression numerical test

[0216] In uniaxial compression numerical tests, there is contact between particles and the wall, requiring μ...bw Also take μ into account bw Set to 0.1, the remaining 8 parameters (particle-particle linear effective modulus E) b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio Particle-to-particle friction coefficient μ, parallel bond tensile strength σ t Parallel bond shear strength τ c internal friction angle As variable parameters in uniaxial compression numerical experiments, the upper and lower limits of eight parameters were determined. The values ​​of the four levels for each experimental factor were determined, as shown in Table 1. Using orthogonal experimental design, the values ​​of L, as shown in Table 2, were selected. 96 4 8 Numerical tests on uniaxial compression of red bed soft rock were conducted using an orthogonal array with 8 factors, 4 levels, and 96 trials.

[0217] Table 1

[0218]

[0219] Table 2

[0220]

[0221] The uniaxial compression modeling process is as follows: Particles are generated inside a cylindrical container with L = 100 mm and W = 50 mm, where the particle size follows a Gaussian normal distribution, R... max / R min The contact parameters were set to 1.66 mm, R to 0.8 mm, and the particle-particle and particle-wall contacts were set as linear model contacts to simulate the rock's occurrence conditions. The contact parameters in the linear model were modified based on experimental parameters, and the particle-particle contact was set as a linear parallel bond model contact. A constant velocity v was applied to the upper and lower loading plates. z To ensure the stability of the model, v z It needs to be kept to a small value.

[0222] 3) Use correlation analysis to identify the microscopic parameters that have the greatest influence on the macroscopic characteristics of rocks.

[0223] Correlation analysis of uniaxial compression numerical test results of 96 groups of red-bed soft rock based on GRA, Pearson, and MI revealed that some microscopic parameters have an impact on the macroscopic characteristic compressive strength σ. f The elastic modulus E and Poisson's ratio v have a significant impact, and the micro-parameters are related to σ. f The maximum correlation coefficient between E and σ is greater than 0.7, while the maximum correlation coefficient between σ and v is 0.56. This indicates that in the material strength deformation characteristics, the mesoscopic parameters mainly affect the material's E and σ. fFurthermore, the correlation results from different algorithms show some discrepancies; therefore, averaging the results yields the detailed parameters and σ. f The degree of correlation between E and v.

[0224] The gamma correlation between contact parameters and macroscopic properties under four different algorithms is as follows: Figure 3 As shown in (a)-(c). Sort the averaged γ from largest to smallest to find the micro-parameters that have a greater impact on E. k rb2 and ; for σ f The microscopic parameters that have a significant impact are μ and σ. t τ c and The micro-parameters that have a significant impact on v are: E b2 k rb2 and Among them, parameters that have a significant impact on E also affect v.

[0225] To further investigate the significant influence of microscopic parameters and σ f The influence mechanism between E and v was studied using the controlled variable method. When a specific parameter with a significant impact was changed, the other parameters all took their minimum (factor level 1), median (factor levels 2 and 3), or maximum (factor level 4) values. The results are as follows: Figure 5 As shown. Figure 5 (a)~(h) represent the relationship between E and the microstructure that has a significant impact on E, where, The slope of the curve is linearly related to E, with a slope of approximately 1.1, while the intercept is controlled by the other parameters; and E varies with k. rb2 and The increase shows a slight decreasing trend, indicating that k rb2 and The effect on E is not significant.

[0226] Analysis of the influence trend of macroscopic characteristics of red bed soft rock based on three types of correlation analysis algorithms and microscopic parameters, such as Figure 4 As shown, the key micro-parameters affecting macroscopic feature E are: This affects the compressive strength σ f and tensile strength σ t The key micro-parameter is the bond shear strength τ. c The key micro-parameters affecting v are and k rb2 In addition, due to and k rb2 The influence on v follows the same trend, therefore it can be... =k rb2 .

[0227] In summary, through a series of analyses of the correlation mechanism between micro-parameters and macro-characteristics, the main micro-parameters affecting the elastic modulus E and Poisson's ratio v are identified as follows: and Both show a linear relationship; while the compressive strength σ f、 Tensile strength σ t The main microscopic parameter is τ c , τ c Then it can be determined by k.

[0228] 4) Predict E and σ using typical ML algorithms. f σ t The PSO algorithm is introduced to optimize the hyperparameters of the model, and the ML model is trained based on the optimal hyperparameters.

[0229] By analyzing data on soft rocks with different microscopic parameters and key contact parameters, a system was established. , and τ c Dataset. The dataset contains various key contact parameter combinations. , and τ c Relationship such as Figure 6 As shown.

[0230] The dataset is divided into training set T and training set T in a 7:3 ratio. train and test set T test .

[0231] The fitting results of the predicted and measured values ​​of each ML model on the training set are as follows: Figure 7 As shown in (a)-(d), the three ML models perform well on the training set. , and τ c The measured values ​​all showed good fitting results, generally fluctuating around the 45° central axis, with most points concentrated within the 10% error threshold. Only a few points were outside the error threshold, indicating that the fitting errors of each ML model were low. Figure 8 From (a) to (d), we can see that XGBoost has the highest training accuracy among the three ML models.

[0232] 5) Evaluate each ML model by combining its prediction accuracy and error, and obtain the optimal ML model.

[0233] The scatter plots of predictions for each ML model on the test set are as follows: Figure 9 As shown in (a)-(d), all three ML models performed well on the test set, generally fluctuating around the 45° central axis. Among them, the XGBoost model was most concentrated around the central axis, followed by the BPNN model, and finally the SVR model.

[0234] The prediction accuracy and error evaluation results of each ML model on the test set are as follows: Figure 10 (a)-(d)- Figure 12 As shown in (a)-(d), on the training set, the MSE, RMSE, and MAE of each ML model are relatively small. Among them, the RSE of each ML model with elastic modulus and k is relatively small. 2 All values ​​are greater than 0.9, demonstrating high fitting accuracy. Figure 10 (a)-(d)- Figure 12 From (a) to (d), we can see that the MAE, RMSE, and MSE of the BPNN and XGBoost models for elastic modulus are relatively small. For compressive strength, the XGBoost model has the smallest MAE, the BPNN model has the smallest RMSE and MAE, and the SVR model has relatively large MAE, RMSE, and MSE. For tensile strength, the XGBoost model has the smallest MAE, RMSE, and MSE, followed by the BPNN model, and finally the SVR model. Therefore, the analysis shows that among the three ML models, XGBoost has the highest testing accuracy.

[0235] Based on the comprehensive evaluation of prediction accuracy and error of ML models, the XGBoost model is the best, therefore the XGBoost model is selected as the optimal ML model.

[0236] The present invention also provides a memory that stores multiple instructions for implementing the method as described in Embodiment 1.

[0237] like Figure 13 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301. The memory 302 stores a plurality of instructions, which can be loaded and executed by the processor to enable the processor to perform the method as described in Embodiment 1.

[0238] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A discrete element contact parameter prediction method based on feature analysis-machine learning, characterized in that, Discrete element contact parameter prediction is performed based on the optimal prediction model for key contact parameters of a linear parallel bonding model. The method includes: S1. Based on the characteristics of soft rock particles and the material parameters of the mold, select the discrete element linear parallel bond model for contact parameter calibration, and determine the model parameters and the range of model parameter values ​​for the discrete element linear parallel bond model to be calibrated. S2. Contact parameters of the discrete element linear parallel bond model are obtained based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parameters affecting the macroscopic characteristics of soft rock. The main micro-parameters are used to construct a low-fidelity dataset of contact parameters and macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation. S3. Based on grey relational analysis (GRA), Pearson correlation analysis, and mutual information (MI), the key contact parameters affecting the macroscopic characteristics of soft rock were determined. S4, a high-fidelity dataset of contact parameters and macroscopic features obtained from a large number of uniaxial compression and Brazilian splitting experiments; S5, a key contact parameter prediction model for discrete element linear parallel bonding model based on machine learning; S6, Determine the optimal prediction model for key contact parameters of the linear parallel bond model; S2 includes: S21, Obtain the first contact parameter - soft rock macroscopic features low-fidelity dataset corresponding to the discrete element linear parallel bonding model based on uniaxial compression test numerical simulation; S22, Obtain the low-fidelity dataset of the second contact parameters-macroscopic features corresponding to the discrete element linear parallel bonding model based on the Brazilian splitting test numerical simulation; among which, the tensile strength σ is selected. t As an indicator of the Brazilian splitting test; S23, the first contact parameter-macro feature low-fidelity dataset and the second contact parameter-macro feature low-fidelity dataset are concatenated to form the contact parameter-soft rock macro feature low-fidelity dataset; S21 includes: (1) Determine the parameters required for the uniaxial compression test, including: Geometric parameters include: axial length L of the specimen, width W of the specimen, porosity n, radius R of the specimen, length-to-radius ratio L / R of the specimen, and maximum size R of the particle or pore feature unit. max and the minimum size R of the particle or pore characteristic unit max The ratio of the maximum to minimum size of a particle or pore feature unit, Rmax / Rmin; Material and deformation parameters, including: particle density ρb, particle-to-particle linear effective modulus E b2 Effective modulus of particle-particle bonding Particle-to-particle stiffness ratio k rb2 Particle-to-particle bond effective stiffness ratio and particle-particle friction coefficient μ; Strength parameters, including: parallel bond tensile strength σ t Parallel bond shear strength τ c and internal friction angle ; The coefficient of friction μ between particles and the wall bw ; S5 includes: S51, based on elastic modulus Compressive strength Tensile strength σ t The tensile-compression ratio k is used to construct key contact parameters, including the effective modulus of particle-particle bonding. Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c The dataset; S52, the high-fidelity dataset of contact parameters-macroscopic features is divided into training set T according to a certain ratio. train and test set T test Wherein, the training set T train Used to build machine learning models and test sets T test Used to test the performance of machine learning models; S53, establish a system based on training set T train Machine learning models, including: (1) Predicting key contact parameter particle-particle bonding effective modulus based on typical machine learning prediction algorithms Particle-to-particle bond stiffness ratio and parallel bond shear strength τ c A machine learning model is established; wherein, the typical machine learning prediction algorithms include backpropagation neural network, support vector regression machine and XGBoost; (2) The hyperparameters of the discrete element linear parallel bonding model are optimized based on the particle swarm optimization algorithm to obtain the optimal hyperparameters; (3) The training set T train Input the machine learning model and train the machine learning model based on the optimal hyperparameters.

2. The discrete element contact parameter prediction method based on feature analysis-machine learning according to claim 1, characterized in that, S3 includes: S31, Grey Relational Analysis (GRA) is used to determine the microscopic parameters that have the most influence on the macroscopic characteristics of soft rock. Based on Grey Relational Analysis (GRA), Pearson correlation analysis, and mutual information (MI), key contact parameters are initially determined from the most influential microscopic parameters, including: (1) The micro-parameters are quantified using correlation metrics through grey relational analysis (GRA), including: A. Determine the analytical sequence; the determined analytical sequence includes: Each macroscopic characteristic value of each uniaxial compression numerical test sample is used as the reference sequence for that uniaxial compression numerical test sample. The contact parameter values ​​of the discrete element linear parallel bond model of each uniaxial compression numerical test sample are used as the comparison sequence of that uniaxial compression numerical test sample. B. Calculate and analyze the correlation coefficients and correlation degrees of various detailed parameters in the sequence; including: Based on the reference and comparison sequences of all uniaxial compression numerical test samples, the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for each uniaxial compression numerical test sample are calculated. Based on the correlation coefficients between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model for all uniaxial compression numerical test samples, the correlation degree between each contact parameter and each macroscopic feature of the discrete element linear parallel bond model is calculated. C. Compressive strength Elastic modulus Poisson's ratio Ranking of influencing factors and identification of the most influential micro-parameters; (2) Based on Pearson correlation analysis and mutual information (MI), the correlation between the most influential microscopic parameters and macroscopic features was analyzed, and the correlation between the most influential microscopic parameters and macroscopic features of soft rock was obtained by comparing the results of grey relational analysis (GRA). Among them, Pearson is used to measure the degree of linear correlation between two variables. The calculation formula of Pearson correlation analysis is as shown in equation (6), and the calculation formula of mutual information (MI) is as shown in equation (7). （6）； In equation (6), For the discrete element linear parallel bonding model, the first The contact parameter and the first Pearson correlation coefficients of macroscopic features for In the first uniaxial compression numerical test sample, the discrete element linear parallel bond model of the first... The average value of each contact parameter for In the uniaxial compression numerical test sample, the first The average of several macroscopic characteristics; （7）； In equation (7), For the discrete element linear parallel bonding model, the first The contact parameter and the first The mutual information value of a macroscopic feature, for and The joint probability mass function, For the discrete element linear parallel bonding model, the first Macro characteristics All possible values, For the discrete element linear parallel bonding model, the first Contact parameters All possible values, The first discrete-element linear parallel bond model represents the... One macro characteristic, The first discrete-element linear parallel bond model represents the... One contact parameter, The first discrete-element linear parallel bond model represents the... Contact parameters Values The probability, i.e. Marginal probability mass function, The first discrete-element linear parallel bond model represents the... Macro characteristics Values The probability, i.e. The marginal probability mass function; (3) Based on grey relational analysis (GRA), Pearson analysis, and mutual information analysis (MI), correlation analysis was performed on the results of uniaxial compression numerical experiments; S32, based on impact analysis, the key contact parameters affecting the macroscopic characteristics of soft rock were preliminarily determined, thereby identifying the main mesoscopic parameters affecting the elastic modulus E and Poisson's ratio v. and Both show a linear relationship; affecting compressive strength σ f and tensile strength σ t The main microscopic parameter is the parallel bond shear strength τ. c Parallel bond shear strength τ c It can be determined by the tension-compression ratio k.

3. The discrete element contact parameter prediction method based on feature analysis-machine learning according to claim 2, characterized in that, S4 includes: S41, the first high-fidelity dataset was obtained based on a large number of uniaxial compression numerical experiments; S42, based on a large number of Brazilian splitting experiments to obtain a second high-fidelity dataset; S43, obtain a high-fidelity dataset of contact parameters and macroscopic features based on the first high-fidelity dataset and the second high-fidelity dataset.

4. The discrete element contact parameter prediction method based on feature analysis-machine learning according to claim 3, characterized in that, S6 includes: S61, calculate the prediction accuracy and error of the machine learning model; Calculate the goodness of fit R 2 Mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE) are used as the prediction accuracy and error metrics of the machine learning model; where: The goodness of fit R 2 The formula is shown in equation (9) below: （9） The mean square error (MSE) is expressed in the following formula (10): (10) The root mean square error (RMSE) formula is shown in equation (11): (11) The formula for the mean absolute error (MAE) is shown in equation (12): (12) In equations (9)-(12), For the sample size, for , and τ c The measured value, For Ē b and k̄ rb τ c The predicted value, for , and τ c The average of the measured values; S62, evaluate the quality of the machine learning model based on the prediction accuracy and error hierarchy, thereby determining the optimal prediction model for the key contact parameters of the linear parallel bonding model; wherein the characteristics of each index are: R 2 The range is between 0 and 1, and the closer the value is to 1, the higher the precision; MSE, RMSE and MAE are all greater than 0, and the closer the value is to 0, the smaller the error.

5. A vibration suppression system for an unmanned aerial vehicle (UAV) used for track interlayer inspection, for implementing the method described in any one of claims 1-4, characterized in that, The system includes: The model and model parameter determination module (101) is used to select the discrete element linear parallel bond model for contact parameter calibration based on the characteristics of soft rock particles and the material parameters of the mold, and to determine the model parameters and the range of model parameter values ​​of the discrete element linear parallel bond model to be calibrated. The contact parameter test acquisition module (102) is used to acquire the contact parameters of the discrete element linear parallel bond model based on a small number of uniaxial compression and Brazilian splitting tests. The contact parameters are the main micro-parallel parameters that affect the macroscopic characteristics of soft rock, and the main micro-parallel parameters are used to form a low-fidelity dataset of contact parameters-macroscopic characteristics of soft rock. The macroscopic characteristics of soft rock include strength and deformation. The critical contact parameter determination module (103) is used to determine the critical contact parameters affecting the macroscopic characteristics of soft rock based on grey relational analysis (GRA), Pearson correlation analysis and mutual information (MI); A high-fidelity dataset acquisition module (104) is used to acquire a high-fidelity dataset of contact parameters and macroscopic features based on a large number of uniaxial compression and Brazilian splitting experiments. The machine learning model building module (105) is used to build a key contact parameter prediction model for a discrete element linear parallel bonding model based on machine learning. The optimal prediction model determination module (106) is used to determine the optimal prediction model for key contact parameters of the linear parallel bonding model.

6. An electronic device, characterized in that, It includes a processor and a memory, the memory storing multiple instructions, and the processor being used to read the instructions and execute the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions, which can be read by a processor and executed as described in any one of claims 1-4.

Images