Block discrete element method parameter determination method, device, equipment and medium

Abstract

The invention discloses a block discrete element method parameter determination method, device and equipment and a medium, and relates to the technical field of rock mechanical behavior simulation, the method comprises the following steps: determining a plurality of parameter value combinations based on each block and the value range of contact parameters, and for each parameter value combination, taking the parameter value combination as input to determine a block discrete element method parameter; and determining a predicted value of the macro-mechanical parameter by using the trained macro-mechanical parameter prediction model, calculating a deviation between the predicted value of the macro-mechanical parameter and a first experimental value of the macro-mechanical parameter to obtain a loss value corresponding to the parameter value combination, and selecting the parameter value combination with the minimum loss value as a final parameter value combination. According to the method, the problems of poor authenticity, low efficiency and low error-tolerant rate of a traditional block discrete element method parameter determination method can be solved, and simulation of rock mechanical behaviors is facilitated.

Description

Technical Field

[0001] The present application relates to the technical field of rock mechanical behavior simulation, and in particular to a block discrete element method parameter determination method, device, equipment and medium. Background Art

[0002] Due to the invisibility of internal rock fractures, people usually need to use numerical simulation to reproduce their mechanical behavior, thereby providing a basis for the prevention and control of rock engineering geological disasters.

[0003] Under the long geological action, rocks usually include a large number of discontinuous surfaces, which greatly affect the physical and mechanical properties of rocks and make it very difficult to simulate the mechanical behavior of rocks. In view of this complex characteristic of rocks, the block discrete element method regards rocks as discontinuous media, that is, it discretizes rocks into rigid or deformable polygonal block aggregates. The blocks are bonded to each other through contact, and the kinematic formulas of contact and blocks are calculated to simulate the mechanical behavior of rocks. Since the block discrete element method can simulate rock fracture, nonlinear large deformation and other behaviors more realistically, it has been widely used in rock engineering simulation.

[0004] In the block discrete element method, the simulation of rock mechanical behavior essentially depends on the block and contact parameters, that is, the properties of the simulated minerals and mineral boundaries. However, the block and contact parameters are microscopic parameters, and current technical means cannot directly measure the block and contact parameters through experiments, because experiments can only measure macroscopic parameters, and it is also impossible to indirectly deduce and calculate the block and contact parameters through other measurable physical parameters of materials. Therefore, when applying the block discrete element method, accurate block and contact parameters are the prerequisite for reproducing rock mechanical behavior.

[0005] When using the block discrete element method to simulate rock mechanical behavior, since the block and contact parameters cannot be directly obtained, the determination of the block and contact parameters usually relies on the trial and error method, which repeatedly adjusts the block and contact parameters until the simulated macromechanical parameters are consistent with the macromechanical parameters of the indoor experiment. Furthermore, the fitting method is derived from the trial and error method, that is, based on a large number of trials and errors, a fitting curve is established to characterize the relationship between the block and contact parameters and the simulated macromechanical parameters, and the block and contact parameters that meet the macromechanical parameter requirements of the indoor experiment are determined by fitting the curve.

[0006] However, there are many problems with the trial and error method and the fitting method in practical applications, as follows.

[0007] (1) Poor authenticity. The trial and error method relies on the process of gradually adjusting the block and contact parameters and observing the results, while the fitting method relies on the relationship between the block and contact parameters and the macroscopic mechanical parameters under the local preset gradient (i.e., the interval between two adjacent values). Therefore, it is difficult for both to ensure that the parameter value combination found is the global optimal one, and they can often only achieve the local optimal one. This limitation often leads to a certain deviation between the simulation results and the actual situation, or only meets the corresponding requirements of some block and contact parameters, which greatly reduces the authenticity of the simulation. In addition, since the trial and error method relies on manual adjustment and observation, and the fitting method relies on manual gradient settings, it is highly subjective.

[0008] (2) Low efficiency. Since a large number of block and contact parameters are involved, the trial-and-error method and fitting method usually require a large number of discrete element numerical simulations, which need to be rerun after each parameter adjustment, resulting in high computational costs and time consumption.

[0009] (3) Low error tolerance. When dealing with complex parameter spaces with high dimensions and nonlinear complex relationships, trial and error and fitting methods face huge challenges. Multiple blocks and contact parameters in complex parameter spaces may interact with each other, making it difficult to find a suitable combination of parameter values ​​through simple adjustments. As the complex parameter space increases, trial and error and fitting methods gradually lose their effectiveness, especially when faced with complex physical phenomena. Summary of the invention

[0010] The purpose of this application is to provide a block discrete element method parameter determination method, device, equipment and medium, which can solve the problems of poor authenticity, low efficiency and low fault tolerance in the traditional block discrete element method parameter determination method.

[0011] To achieve the above objectives, this application provides the following solutions.

[0012] In a first aspect, the present application provides a method for determining parameters of a block discrete element method, and the method for determining parameters of a block discrete element method comprises the following steps.

[0013] Based on the value range of each block and contact parameter, multiple parameter value combinations are determined; the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combinations include the values ​​of each block and contact parameter.

[0014] For each of the parameter value combinations, the parameter value combination is taken as input, the predicted value of the macromechanical parameter is determined using a trained macromechanical parameter prediction model, and the deviation between the predicted value of the macromechanical parameter and the first experimental value of the macromechanical parameter is calculated to obtain the loss value corresponding to the parameter value combination; the first experimental value is the value of the macromechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination, the macromechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle.

[0015] The parameter value combination with the smallest loss value is selected as the final parameter value combination.

[0016] In a second aspect, the present application provides a device for determining parameters of a block discrete element method, and the device for determining parameters of a block discrete element method includes the following modules.

[0017] The data acquisition module is used to determine multiple parameter value combinations based on the value range of each block and contact parameter; the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combination includes the value of each block and contact parameter.

[0018] The loss calculation module is used for determining the predicted value of the macro-mechanical parameter for each parameter value combination, taking the parameter value combination as input, using a trained macro-mechanical parameter prediction model, calculating the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter, and obtaining the loss value corresponding to the parameter value combination; the first experimental value is the value of the macro-mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination, the macro-mechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle.

[0019] The parameter determination module is used to select the parameter value combination with the smallest loss value as the final parameter value combination.

[0020] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned block discrete element method parameter determination method.

[0021] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned block discrete element method parameter determination method.

[0022] According to the specific embodiments provided in this application, this application has the following technical effects.

[0023] The present application provides a block discrete element method parameter determination method, device, equipment and medium, based on the value range of each block and contact parameter, multiple parameter value combinations are determined, for each parameter value combination, the parameter value combination is used as input, the predicted value of the macromechanical parameter is determined using a trained macromechanical parameter prediction model, the deviation between the predicted value of the macromechanical parameter and the first experimental value of the macromechanical parameter is calculated, the loss value corresponding to the parameter value combination is obtained, and the parameter value combination with the smallest loss value is selected as the final parameter value combination. The present application can traverse multiple parameter value combinations of block and contact parameters, reduce the subjectivity in parameter selection, solve the problem of true difference, use a trained macromechanical parameter prediction model to predict macromechanical parameters, avoid a large number of discrete element numerical simulations, solve the problem of low efficiency, and can handle complex parameter spaces with high dimensions and nonlinear complex relationships, solve the problem of low fault tolerance, thereby solving the problems of poor authenticity, low efficiency, and low fault tolerance in traditional block discrete element method parameter determination methods, and accurately and efficiently determine the parameters of the block discrete element method. In summary, this application reduces subjectivity and avoids a large number of discrete element numerical simulations by traversing multiple parameter value combinations. At the same time, it can handle complex high-dimensional nonlinear parameter spaces, solve the problems of poor authenticity, low efficiency and low fault tolerance, and thus realize an accurate and efficient block discrete element method parameter determination process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 An application environment diagram of a block discrete element method parameter determination method provided in Example 1 of the present application.

[0026] Figure 2 A flow chart of a block discrete element method parameter determination method provided in Example 1 of the present application.

[0027] Figure 3 A schematic diagram of a technical route for a block discrete element method parameter determination method provided in Example 1 of the present application.

[0028] Figure 4Schematic diagram of the comparison between the numerical simulation results of the Young's modulus of igneous rock provided in Example 1 of the present application and the experimental data.

[0029] Figure 5 Schematic diagram of the comparison between the numerical simulation results of the Poisson's ratio of igneous rock provided in Example 1 of the present application and the experimental data.

[0030] Figure 6 This is a schematic diagram comparing the numerical simulation results of the uniaxial compressive strength of igneous rock provided in Example 1 of the present application with the experimental data.

[0031] Figure 7 This is a schematic diagram showing the comparison between the numerical simulation results and experimental data of the Brazilian tensile strength of igneous rock provided in Example 1 of the present application.

[0032] Figure 8 Schematic diagram comparing the numerical simulation results of the cohesion of igneous rock provided in Example 1 of the present application with the experimental data.

[0033] Fig. 9 Schematic diagram of the comparison between the numerical simulation results and experimental data of the internal friction angle of igneous rock provided in Example 1 of the present application.

[0034] Fig.10 A schematic diagram of the functional modules of a block discrete element method parameter determination device provided in Example 2 of the present application.

[0035] Fig.11 A schematic diagram of the structure of a computer device provided in Example 3 of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] Example 1.

[0038] The block discrete element method parameter determination method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on the cloud or other servers. The terminal can send multiple parameter value combinations of the block and contact parameters to be processed to the server. After the server receives the multiple parameter value combinations of the block and contact parameters to be processed, for each parameter value combination, the server uses the parameter value combination as input, and uses the trained macro-mechanical parameter prediction model to determine the predicted value of the macro-mechanical parameter, calculates the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter, and obtains the loss value corresponding to the parameter value combination; selects the parameter value combination with the smallest loss value as the final parameter value combination. The server can feedback the final parameter value combination to the terminal.

[0039] In addition, in some embodiments, the parameter determination method of the block discrete element method can also be implemented independently by a server or a terminal. For example, the terminal can directly process multiple parameter value combinations of the block and contact parameters to be processed, or the server can obtain multiple parameter value combinations of the block and contact parameters to be processed from a data storage system and process multiple parameter value combinations of the block and contact parameters to be processed.

[0040] The terminals may be, but are not limited to, various desktop computers, laptops, smart phones, tablet computers, IoT devices and portable wearable devices. IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server may be implemented as an independent server or a server cluster consisting of multiple servers, or a cloud server.

[0041] In an exemplary embodiment, Figure 2 As shown, a block discrete element method parameter determination method is provided, which is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The following steps are used to illustrate the server in the example.

[0042] Step S1, based on the value range of each block and contact parameter, determine multiple parameter value combinations; the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combination includes the value of each block and contact parameter.

[0043] Step S2, for each of the parameter value combinations, take the parameter value combination as input, use the trained macro-mechanical parameter prediction model to determine the predicted value of the macro-mechanical parameter, calculate the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter, and obtain the loss value corresponding to the parameter value combination; the first experimental value is the value of the macro-mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination, the macro-mechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle.

[0044] Step S3, selecting the parameter value combination with the smallest loss value as the final parameter value combination.

[0045] In order to determine the block and contact parameters, most traditional technical means such as trial and error method and fitting method are currently used. However, such traditional technical means have the problems of poor authenticity, low efficiency and low fault tolerance, which brings great challenges to the rapid and accurate determination of block and contact parameters. In recent years, artificial intelligence and machine learning technologies have shown great potential in parameter optimization and uncertainty analysis, especially in dealing with complex, nonlinear relationships and high-dimensional parameter space. Therefore, this embodiment uses machine learning to quickly and accurately determine the block and contact parameters in the block discrete element method. Specifically, a block discrete element method parameter determination method based on machine learning is proposed to solve the problems of poor authenticity, low efficiency and low fault tolerance in the traditional block discrete element method parameter determination method. The implementation of the above steps S1 to S3 can solve the problems of poor authenticity, low efficiency and low fault tolerance in the traditional block discrete element method parameter determination method, and accurately and quickly determine the parameters of the block discrete element method, thereby serving rock mechanics experiments and rock engineering geological disaster prevention and control simulations.

[0046] This embodiment obtains input parameters and output parameters through a numerical simulation tool based on the block discrete element method to construct a data set, and constructs a multi-scale pre-trained model framework including a random forest model (abbreviated as RF), a Gaussian process regression model (abbreviated as GPR), a support vector regression model (abbreviated as SVR) and a deep neural network model (abbreviated as DNN). The multi-scale pre-trained model framework is trained using the data set to obtain a trained macroscopic mechanical parameter prediction model. On the basis of the trained macroscopic mechanical parameter prediction model, the grid search method is used to perform parameter inversion, specifically define the value range and create a parameter network (i.e., determine multiple parameter value combinations), and determine the optimal parameter value combination of the block discrete element method through the minimum loss value. Finally, the simulation results and experimental results are used for verification. Figure 3 As shown, the specific steps include the following steps.

[0047] 1.1 Collection and preparation of data sets.

[0048] According to the characteristics of rock materials and previous studies, the range and gradient of the microscopic block and contact parameters (i.e., the interval between two adjacent values) are reasonably set to generate multiple parameter value combinations. The parameter value combinations include the values ​​of each block and contact parameter. The values ​​of each block and contact parameter in different parameter value combinations are not exactly the same. The block and contact parameters can include deformation-related parameters and strength-related parameters. The deformation-related parameters can include the contact normal stiffness. k n , contact shear stiffness k s , the ratio of the contact normal stiffness to the contact shear stiffness k n / k s , Young's modulus of rock E b and the rock Poisson's ratio ν b , strength-related parameters can include contact cohesion c cont , contact internal friction angle cont and contact tensile strength σ t cont . A numerical simulation model of rock samples is established in the discrete element method software, the geometric shape and boundary conditions of the rock samples are specifically defined, and the initial size of the rock particles is set. The initial size can be 2 mm, 3 mm, and 4 mm, etc., which can be set according to user needs. The size division is to enhance the universality of the results and build a corresponding numerical simulation model. According to the experimental design plan, different parameter value combinations are input into the numerical simulation model, and then the numerical simulation model is run to record the simulation results of the rock samples under each parameter value combination to obtain the simulation values ​​of the macroscopic mechanical parameters. The macroscopic mechanical parameters can include rock deformation characteristics and rock strength characteristics. The rock deformation characteristics can include Young's modulus. E and Poisson's ratio μ Rock strength properties can include uniaxial compressive strength UCS, Brazilian tensile strength BTS, cohesion c and internal friction angle The rationality of the simulation results is checked, and the simulation results that are abnormal or inconsistent with physical laws are excluded. Finally, the preset parameter value combination in the numerical simulation and the simulated values ​​of the macroscopic mechanical parameters obtained are used as the input parameters and output parameters of the data set to obtain the data set.

[0049] (ii) Establish four different machine learning models.

[0050] (1) Divide the data set into a training set and a test set, ensuring that the data in each subset is evenly distributed. The main function of the training set is to allow the model to learn the patterns and features of the data, that is, to train the model, and the test set is used to evaluate the model performance.

[0051] (2) Determine whether the four machine learning models, namely, random forest model, Gaussian process regression model, support vector regression model and deep neural network model, have been established.

[0052] (3) If not, return to step (2) and build each machine learning model for training.

[0053] (4) If yes, proceed to the next step.

[0054] (5) Perform hyperparameter optimization. Specifically, use the grid search method to determine multiple hyperparameter combinations. For each hyperparameter combination, split the training set into a training subset and a validation subset. Use the training subset for training, and test the training effect on the validation subset (which can be characterized by accuracy). Finally, select the optimal hyperparameter combination based on the training effect. The hyperparameter optimization method mainly uses GridSearchCV and RandomizedSearchCV. This method automatically performs cross-validation. By default, 5-fold cross-validation is used. Depending on the size of the data set, this embodiment can select 5-fold cross-validation or 10-fold cross-validation.

[0055] (6) Determine whether each machine learning model using the optimal hyperparameter combination has achieved the training goal. The training goal can be to reach the maximum number of iterations or the loss value is less than the preset value.

[0056] (7) If not, return to step (6).

[0057] (8) If yes, then proceed to the next stage and save the trained machine learning model to obtain a trained macro-mechanical parameter prediction model. At the same time, the importance of features can be analyzed based on the trained macro-mechanical parameter prediction model to explore how block and contact parameters affect macro-mechanical parameters. For example, in the random forest model, the importance ranking of features can be directly viewed.

[0058] At this time, in this embodiment, before taking the parameter value combination as input and using the trained macro-mechanical parameter prediction model to determine the predicted value of the macro-mechanical parameter, the block discrete element method parameter determination method of this embodiment also includes: obtaining a data set, the data set includes multiple samples and a label corresponding to each sample, the sample is a parameter value combination of block and contact parameters, the label is a first simulated value of the macro-mechanical parameter, the first simulated value is the value of the macro-mechanical parameter obtained by simulating the mechanical behavior of the rock sample under the sample using the block discrete element method; establishing multiple initial prediction models, the initial prediction model is a machine learning model; for each initial prediction model, the initial prediction model is trained using the data set to obtain a trained model and the prediction performance of the trained model; and the trained model with the best prediction performance is selected as the trained macro-mechanical parameter prediction model.

[0059] Among them, the initial prediction models include random forest model, Gaussian process regression model, support vector regression model and deep neural network model.

[0060] Among them, the initial prediction model is trained using the data set to obtain the trained model and the prediction performance of the trained model, specifically including: dividing the data set into a training set and a test set, using the training set to train the initial prediction model to obtain a trained model, using the test set to determine the prediction performance of the trained model, if the prediction performance does not meet the requirements, then using the trained model as the initial prediction model for the next iteration, returning to the step of "training the initial prediction model using the training set", if the prediction performance meets the requirements, then using the trained model as the trained model, and using the test set to determine the prediction performance of the trained model.

[0061] The prediction performance can be characterized by accuracy, and the prediction performance meeting the requirement means that the prediction performance is greater than a preset value.

[0062] (iii) Parameter inversion via grid search method.

[0063] (1) According to the value ranges of block and contact parameters in the data set, determine the reasonable value range of each block and contact parameter, and set the step size, that is, the size of each parameter increment in the grid search, to facilitate the subsequent generation of parameter value combinations.

[0064] (2) Based on the defined value range and step size, determine whether all possible parameter value combinations have been generated.

[0065] (3) If not, return to step (2).

[0066] (4) If yes, proceed to the next step.

[0067] (5) Initialize the minimum loss value to a larger value and proceed to the next step.

[0068] (6) The current parameter value combination is input into the trained macro-mechanical parameter prediction model to predict the macro-mechanical parameters of the rock sample. According to the deviation between the prediction result of the trained macro-mechanical parameter prediction model and the actual experimental data, the loss value of the current parameter value combination is calculated. The loss value is obtained by summing the square of the difference between each predicted value and the experimental value.

[0069] (7) Determine whether the loss value of the current parameter value combination is less than the recorded minimum loss value.

[0070] (8) If not, skip the update and proceed to the next parameter value combination.

[0071] (9) If so, update the minimum loss value and set the current parameter value combination as the optimal parameter value combination.

[0072] (10) Determine whether all parameter value combinations have been traversed.

[0073] (11) If not, select a parameter value combination that has not been traversed as the current parameter value combination and return to step (6).

[0074] (12) If so, the grid search ends and the next stage is entered. Finally, the parameter value combination with the smallest loss value is recorded and output. This parameter value combination represents the values ​​of the block and contact parameters that make the macroscopic mechanical parameters predicted by the model closest to the experimental data.

[0075] At this time, in this embodiment, based on the value range of each block and contact parameter, multiple parameter value combinations are determined, the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combination includes the value of each block and contact parameter; for each parameter value combination, the parameter value combination is used as input, and the trained macroscopic mechanical parameter prediction model is used to determine the predicted value of the macroscopic mechanical parameter, and the deviation between the predicted value of the macroscopic mechanical parameter and the first experimental value of the macroscopic mechanical parameter is calculated to obtain the loss value corresponding to the parameter value combination, and the first experimental value is the value of the macroscopic mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination; the parameter value combination with the smallest loss value is selected as the final parameter value combination.

[0076] Among them, block and contact parameters include deformation-related parameters and strength-related parameters. Deformation-related parameters include contact normal stiffness, contact shear stiffness, the ratio of contact normal stiffness to contact shear stiffness, rock Young's modulus and rock Poisson's ratio. Strength-related parameters include contact cohesion, contact internal friction angle and contact tensile strength. Macroscopic mechanical parameters include rock deformation characteristics and rock strength characteristics. Rock deformation characteristics include Young's modulus and Poisson's ratio. Rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle.

[0077] Among them, based on the value range of each block and contact parameter, multiple parameter value combinations are determined, specifically including: obtaining the value range and step size of each block and contact parameter; for each block and contact parameter, multiple values ​​are taken within the value range based on the step size to obtain multiple values ​​of the block and contact parameters, that is, based on the value range and step size, multiple values ​​are generated within the value range, specifically starting from the lower limit of the value range, and gradually increasing the value based on the step size until the upper limit of the value range is reached; multiple values ​​of all block and contact parameters are randomly combined to obtain multiple parameter value combinations of block and contact parameters.

[0078] Among them, the deviation between the predicted value of the macroscopic mechanical parameter and the first experimental value of the macroscopic mechanical parameter is calculated to obtain the loss value corresponding to the parameter value combination, specifically including: calculating the difference between the predicted value and the first experimental value of each macroscopic mechanical parameter (including Young's modulus, Poisson's ratio, uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle); calculating the sum of the squares of all the differences, that is, summing the squares of all the differences to obtain the loss value corresponding to the parameter value combination.

[0079] (IV) Simulation and comparative verification using block discrete element method.

[0080] (1) The parameter value combination with the minimum loss value predicted by machine learning is input into the discrete element method software to carry out uniaxial compression numerical simulation, triaxial compression numerical simulation and Brazilian splitting numerical simulation. The numerical simulation model size and boundary conditions are the same as those used for uniaxial compression, triaxial compression and Brazilian splitting specimens and loading conditions, and the simulated values ​​of the macroscopic mechanical parameters are obtained.

[0081] (2) Determine whether the macroscopic mechanical parameters obtained by numerical simulation are equal to those obtained by experiment.

[0082] (3) If not, return to the previous process, either to the step of training the model or to the step of parameter inversion. If the large loss value is caused by poor model performance, you can choose to return to the step of training the model. If the large loss value is caused by improper parameter setting, you can choose to return to the step of parameter inversion, that is, adjust the machine learning model or adjust the value range, re-invert the parameters, and then input the discrete element method software for verification.

[0083] (4) If so, directly output the parameter value combination with the smallest loss value predicted by machine learning as the modeling parameters of the block discrete element method.

[0084] At this time, in this embodiment, the parameter value combination with the smallest loss value is selected as the final parameter value combination, specifically including: selecting the parameter value combination with the smallest loss value as the alternative parameter value combination; obtaining a second simulation value of the macroscopic mechanical parameter and a second experimental value of the macroscopic mechanical parameter corresponding to the alternative parameter value combination, the second simulation value is the value of the macroscopic mechanical parameter obtained by simulating the mechanical behavior of the rock sample using the block discrete element method under the alternative parameter value combination, and the second experimental value is the value of the macroscopic mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the alternative parameter value combination. value; determine whether the second simulation value of the macro-mechanical parameter is equal to the second experimental value of the macro-mechanical parameter; if so, use the alternative parameter value combination as the final parameter value combination; if not, adjust the value range of each block and contact parameter, and return to the step of "determining multiple parameter value combinations based on the value range of each block and contact parameter", or update the trained macro-mechanical parameter prediction model, and return to the step of "for each parameter value combination, take the parameter value combination as input, and use the trained macro-mechanical parameter prediction model to determine the predicted value of the macro-mechanical parameter".

[0085] The method of this embodiment has wide applicability and is applicable to multiple block discrete element method software such as UDEC (Universal Distinct Element Code) and 3DEC (3-Dimension Distinct Element Code), and is applicable to any size within a reasonable range, and can be applied to different lithological materials, such as igneous rock, sedimentary rock and metamorphic rock. In the following, this embodiment specifically uses UDEC software, selects 2 mm, 3 mm and 4 mm and igneous rock as examples for analysis and explanation, and uses igneous rock as a representative for result analysis, but also fully considers the results of other lithologies to ensure the universality and accuracy of the method.

[0086] The parameter space includes deformation-related parameters and strength-related parameters. Reasonable gradients are set and corresponding parameter value combinations are generated. UDEC software is used to carry out numerical simulations of uniaxial compression, triaxial compression and Brazilian splitting. The width of the compression numerical simulation model is 50 mm and the height is 100 mm. The diameter of the splitting numerical simulation model is 50 mm, which meets the ASTM (American Society of Testing Materials) standard. All numerical simulation models are composed of Voronoi grids formed by deformable polygons bonded by contacts. The contacts follow the classical Coulomb friction law. The numerical simulation models are sandwiched between upper and lower numerical simulation steel plates with a width of 50 mm and a height of 10 mm. The width of the numerical simulation steel plate is 50 mm, the height is 10 mm, the Young's modulus is 140 GPa, the Poisson's ratio is 0.28, and the contact normal stiffness of the interface between the numerical simulation model and the numerical simulation steel plate is k n Set to 10 6 GPa / m, and no contact shear stiffness k s , contact adhesion c cont , contact internal friction angle cont and contact tensile strength σ t cont ,To eliminate the boundary effect, during the simulation, the lower numerical simulation steel plate was fixed, ,while the upper numerical simulation steel plate compressed the numerical simulation model at a constant ,speed of 0.015 m / s.

[0087] Based on the preset parameter value combinations of inter-particle block and contact parameters in numerical simulation and their corresponding macroscopic mechanical parameters, the training and performance evaluation of machine learning models were carried out, and a multi-scale pre-training model framework including random forest model, Gaussian process regression model, support vector regression model and deep neural network model was specifically constructed. The results show that the random forest model exhibits the best prediction performance, and its prediction accuracy remains stable above 0.99. The deep neural network model shows significant advantages in dealing with complex nonlinear relationships, and its prediction accuracy ranges from 0.97 to 0.99, which is particularly suitable for simulating mechanical behaviors with highly nonlinear characteristics. In contrast, the prediction performance of the support vector regression model and the Gaussian process regression model is relatively poor, especially for the 3 mm size, the prediction accuracy of uniaxial compressive strength and cohesion is significantly reduced.

[0088] Taking the deep neural network model as an example, the grid search method is applied to perform parameter inversion, the parameter value range is defined and a parameter network is created, all possible parameter value combinations are input into the trained deep neural network model in turn, the macroscopic mechanical parameters of the rock sample are predicted, and the deviation between the predicted results of the calculation model and the experimental data is finally determined to obtain the parameter value combination that optimizes the model prediction accuracy.

[0089] So far, the optimal parameter value combination predicted by the machine learning model, including contact normal stiffness, contact cohesion, contact internal friction angle, contact tensile strength, etc., has been determined. The optimal parameter value combination of the above simulation is input into the UDEC software to perform uniaxial compression numerical simulation and Brazilian splitting numerical simulation, and the simulation results are compared with the experimental results. Taking igneous rock as an example, Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Fig. 9 The comparative analysis between the numerical simulation results and the experimental data is presented, covering key macroscopic mechanical parameters such as Young's modulus, Poisson's ratio, uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle. The prediction performance of the model is quantitatively evaluated by calculating the statistical error value (mean square error, mean absolute error, coefficient of determination and relative error) of each macroscopic mechanical parameter. Figure 4-Figure 9 The predicted values ​​1, 2, and 3 in the figure correspond to the predicted results of 2 mm, 3 mm, and 4 mm sizes, respectively. l Representative size. The results show that as the preset size increases from 2 mm to 4 mm, the prediction accuracy of most macroscopic mechanical parameters is significantly improved, among which the prediction results of Young's modulus and uniaxial compressive strength are particularly accurate. From the perspective of lithology, the prediction effect of igneous rock is the best, especially at a size of 4 mm. The prediction effect of sedimentary rock is slightly lower than that of igneous rock, and the prediction effect of metamorphic rock is relatively poor, especially in the prediction of internal friction angle and Brazilian tensile strength. There are large deviations. The prediction accuracy of Young's modulus and uniaxial compressive strength is high in all lithologies, and the determination coefficient is close to 1, while the prediction errors of internal friction angle and Brazilian tensile strength are large, especially in sedimentary rocks and metamorphic rocks. In general, the prediction results of the model have high accuracy, can effectively reflect the mechanical properties of different lithological materials, and successfully realize the rapid and accurate determination of block discrete element method parameters.

[0090] This embodiment uses machine learning to establish the relationship between the simulated block and contact parameters and the macroscopic mechanical parameters, and uses the grid search method to perform parameter inversion, which has the following advantages.

[0091] (1) The real block and contact parameters can be accurately obtained. The subjectivity and uncertainty of the trial and error method and the fitting method may miss some optimal solutions. Especially when the parameter space is large, the trial and error method may not be able to fully explore all possible parameter value combinations, thus missing the opportunity for the optimal solution. Machine learning is data-driven and uses a large amount of numerical simulation data to learn potential patterns and laws. In addition, machine learning can dynamically adjust hyperparameters and optimize performance during model training, thereby improving authenticity and traversing all possible parameter value combinations. This method reduces the subjectivity in parameter selection and makes the parameter optimization process more objective and systematic.

[0092] (2) High computational efficiency. Machine learning can quickly find the optimal combination of parameter values ​​through efficient optimization algorithms (such as grid search). Through automated data processing and model training, it can significantly reduce the time and workload of manual parameter adjustment. Compared with traditional trial and error and fitting methods, especially when processing large amounts of simulated data, machine learning can more effectively utilize computing resources. It reduces computing requirements in an intelligent way, greatly saving computing costs.

[0093] (3) High fault tolerance. Machine learning models can handle complex nonlinear relationships between input parameters and output results, which is particularly important for complex physical phenomena in the block discrete element method. Machine learning can handle high-dimensional and complex parameter spaces, and is particularly suitable for situations in the discrete element method where there are many microscopic parameter combinations and complex interactions.

[0094] The present application also provides an application scenario, which applies the above-mentioned block discrete element method parameter determination method. Specifically, the block discrete element method parameter determination method provided in this embodiment can be applied in a rock mechanics behavior simulation scenario. The rock mechanics behavior simulation scenario includes a parameter determination link and a simulation link. The parameter determination link is used to determine the final parameter value combination, and the simulation link is used to simulate the rock mechanics behavior based on the final parameter value combination. The block discrete element method parameter determination method provided in this embodiment belongs to the parameter determination link.

[0095] Example 2.

[0096] Based on the same inventive concept, the embodiment of the present application also provides a block discrete element method parameter determination device for implementing the block discrete element method parameter determination method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more block discrete element method parameter determination device embodiments provided below can refer to the limitations of the block discrete element method parameter determination method above, and will not be repeated here.

[0097] In an exemplary embodiment, Fig.10 As shown, a block discrete element method parameter determination device is provided, and the block discrete element method parameter determination device includes the following modules.

[0098] The data acquisition module M1 is used to determine multiple parameter value combinations based on the value range of each block and contact parameter; the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combination includes the value of each block and contact parameter.

[0099] The loss calculation module M2 is used for determining the predicted value of the macro-mechanical parameter for each parameter value combination, taking the parameter value combination as input, using the trained macro-mechanical parameter prediction model, calculating the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter, and obtaining the loss value corresponding to the parameter value combination; the first experimental value is the value of the macro-mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination, the macro-mechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle.

[0100] The parameter determination module M3 is used to select the parameter value combination with the smallest loss value as the final parameter value combination.

[0101] Example 3.

[0102] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a block discrete element method parameter determination method is implemented.

[0103] Those skilled in the art will understand that Fig.11The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0104] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the block discrete element method parameter determination method in Example 1 is implemented.

[0105] Example 4.

[0106] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the block discrete element method parameter determination method in Example 1.

[0107] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0108] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for determining parameters of a block discrete element method, characterized in that: The block discrete element method parameter determination method comprises: Based on the value range of each block and contact parameter, a plurality of parameter value combinations are determined; the block and contact parameters are modeling parameters in the block discrete element method, and the parameter value combination includes the value of each block and contact parameter; For each of the parameter value combinations, the parameter value combination is used as input, and the predicted value of the macro-mechanical parameter is determined by using a trained macro-mechanical parameter prediction model, and the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter is calculated to obtain the loss value corresponding to the parameter value combination; the first experimental value is the value of the macro-mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the parameter value combination, and the macro-mechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle; The parameter value combination with the smallest loss value is selected as the final parameter value combination.

2. The block discrete element method parameter determination method according to claim 1, characterized in that: The block and contact parameters include deformation-related parameters and strength-related parameters. The deformation-related parameters include contact normal stiffness, contact shear stiffness, the ratio of contact normal stiffness to contact shear stiffness, rock Young's modulus and rock Poisson's ratio. The strength-related parameters include contact cohesion, contact internal friction angle and contact tensile strength.

3. The block discrete element method parameter determination method according to claim 1, characterized in that: Based on the value range of each block and contact parameter, multiple parameter value combinations are determined, including: Get the value range and step size of each block and contact parameter; For each of the block and contact parameters, multiple values ​​are taken within the value range based on the step size to obtain multiple values ​​of the block and contact parameters; Multiple values ​​of all the block and contact parameters are randomly combined to obtain multiple parameter value combinations.

4. The block discrete element method parameter determination method according to claim 1, characterized in that: Before determining the predicted value of the macroscopic mechanical parameter using the trained macroscopic mechanical parameter prediction model with the parameter value combination as input, the block discrete element method parameter determination method further includes: Acquire a data set; the data set includes a plurality of samples and a label corresponding to each of the samples, the sample is a parameter value combination, the label is a first simulated value of a macroscopic mechanical parameter, and the first simulated value is a value of the macroscopic mechanical parameter obtained by simulating the mechanical behavior of the rock sample using a block discrete element method under the sample; Establishing multiple initial prediction models; the initial prediction models are machine learning models; For each of the initial prediction models, the initial prediction model is trained using the data set to obtain a trained model and the prediction performance of the trained model; The trained model with the best prediction performance was selected as the trained macroscopic mechanical parameter prediction model.

5. The block discrete element method parameter determination method according to claim 4, characterized in that: The initial prediction models include a random forest model, a Gaussian process regression model, a support vector regression model and a deep neural network model.

6. The block discrete element method parameter determination method according to claim 1, characterized in that: Calculate the deviation between the predicted value of the macromechanical parameter and the first experimental value of the macromechanical parameter to obtain the loss value corresponding to the parameter value combination, specifically including: calculating the difference between the predicted value and the first experimental value of each macromechanical parameter, and calculating the sum of the squares of all the differences to obtain the loss value corresponding to the parameter value combination.

7. The block discrete element method parameter determination method according to claim 1, characterized in that: The parameter value combination with the smallest loss value is selected as the final parameter value combination, specifically including: Selecting the parameter value combination with the smallest loss value as the candidate parameter value combination; Obtaining a second simulation value of the macroscopic mechanical parameter and a second experimental value of the macroscopic mechanical parameter corresponding to the alternative parameter value combination; the second simulation value is the value of the macroscopic mechanical parameter obtained by simulating the mechanical behavior of the rock sample using the block discrete element method under the alternative parameter value combination, and the second experimental value is the value of the macroscopic mechanical parameter obtained by experimenting on the mechanical behavior of the rock sample under the alternative parameter value combination; determining whether a second simulation value of the macroscopic mechanical parameter is equal to a second experimental value of the macroscopic mechanical parameter; If yes, the candidate parameter value combination is used as the final parameter value combination; If not, adjust the value range of each block and contact parameter and return to the step of "determining multiple parameter value combinations based on the value range of each block and contact parameter", or update the trained macro-mechanical parameter prediction model and return to the step of "for each of the parameter value combinations, take the parameter value combination as input, and use the trained macro-mechanical parameter prediction model to determine the predicted value of the macro-mechanical parameter".

8. A block discrete element method parameter determination device, characterized in that: The block discrete element method parameter determination device comprises: A data acquisition module, for determining a plurality of parameter value combinations based on a value range of each block and contact parameter; the block and contact parameters are modeling parameters in a block discrete element method, and the parameter value combination includes a value of each block and contact parameter; A loss calculation module is used for determining the predicted value of the macro-mechanical parameter using the trained macro-mechanical parameter prediction model for each parameter value combination, taking the parameter value combination as input, calculating the deviation between the predicted value of the macro-mechanical parameter and the first experimental value of the macro-mechanical parameter, and obtaining the loss value corresponding to the parameter value combination; the first experimental value is the value of the macro-mechanical parameter obtained by conducting an experiment on the mechanical behavior of the rock sample under the parameter value combination, the macro-mechanical parameters include rock deformation characteristics and rock strength characteristics, the rock deformation characteristics include Young's modulus and Poisson's ratio, and the rock strength characteristics include uniaxial compressive strength, Brazilian tensile strength, cohesion and internal friction angle; The parameter determination module is used to select the parameter value combination with the smallest loss value as the final parameter value combination.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the block discrete element method parameter determination method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the block discrete element method parameter determination method described in any one of claims 1 to 7 is implemented.

Images