Discrete element parameter calibration method and device and storage medium

Abstract

The invention discloses a discrete element parameter calibration method and device and a storage medium, and the method comprises the steps: obtaining a current calibration demand, and initializing a calibration parameter space based on the current calibration demand; a target agent model is determined in an agent model pool according to the calibration parameter space, the calibration parameter space is predicted through the target agent model, candidate calibration parameters are obtained, and a plurality of types of agent models obtained based on discrete element simulation data set training are stored in the agent model pool; performing discrete element simulation on the candidate calibration parameters to obtain a current simulation physical quantity; and when the current simulation physical quantity does not meet the current calibration requirement, executing a preset discrete element simulation iterative optimization algorithm by taking the candidate calibration parameter as an iteration initial point to obtain a target calibration parameter. Therefore, consumption of computing resources is effectively reduced on the premise that calibration precision is guaranteed, and overall efficiency and scene adaptability of parameter calibration are improved.

Description

Technical Field

[0001] This application relates to the field of parameter calibration technology, and in particular to a discrete element parameter calibration method, device and storage medium. Background Technology

[0002] The Discrete Element Method (DEM) is a mainstream numerical method for simulating the dynamic behavior of particulate systems. It is widely used in industrial equipment simulations such as powder mixing, material conveying, and particle erosion in fields like chemical engineering, mining, and agriculture. With the improvement of computer performance and the increasing demands for accuracy in engineering simulations, the role of the DEM in process design and equipment optimization is becoming increasingly prominent. However, the accuracy of DEM simulations highly depends on the reasonable setting of particle physical properties and contact model parameters. These parameters are often difficult to obtain through direct measurement and require parameter calibration techniques to match the simulation results with physical experimental observations. Parameter calibration has become a crucial step in ensuring the reliability of DEM simulations in engineering.

[0003] Existing discrete element method (DEM) parameter calibration methods mainly rely on manual trial and error or a single simulation traversal of the parameter space. When dealing with multi-parameter coupled problems, the computational load increases exponentially with the parameter dimension, resulting in excessively long calibration times. Furthermore, repetitive calculations for similar materials or similar parameter ranges are common in different calibration projects. Numerous already calculated parameter combinations are repeatedly executed in different tasks, leading to a significant waste of computational resources and severely limiting the practical efficiency of DEM in iterative engineering design. Summary of the Invention

[0004] The main objective of this application is to provide a discrete element parameter calibration method, device, and storage medium, aiming to solve the technical problem of low efficiency in existing discrete element parameter calibration methods.

[0005] To achieve the above objectives, this application proposes a discrete element parameter calibration method, the method comprising: Obtain the current calibration requirements and initialize the calibration parameter space based on the current calibration requirements; The target proxy model is determined in the proxy model pool based on the calibration parameter space, and the calibration parameter space is predicted by the target proxy model to obtain candidate calibration parameters. The proxy model pool stores several categories of proxy models trained based on the discrete element simulation dataset. Discrete element simulation is performed on the candidate calibration parameters to obtain the current simulated physical quantities; When the current simulated physical quantity does not meet the current calibration requirements, the candidate calibration parameters are used as the initial point of iteration to execute a preset discrete element simulation iterative optimization algorithm to obtain the target calibration parameters.

[0006] In one embodiment, the current calibration requirement further includes: a calibration parameter range; before the step of determining the target proxy model in the proxy model pool based on the calibration parameter space, it further includes: Query the discrete element parameter simulation database to see if there is a historical dataset that meets the calibration parameter range; If it exists, the historical dataset that meets the calibration parameter range is determined as the discrete element simulation dataset, which includes several sample calibration parameters and corresponding sample simulation physical quantities. The initial agent models of different categories are trained using the discrete element simulation dataset, and the agent model pool is constructed based on the training results.

[0007] In one embodiment, after the step of querying the discrete element parameter simulation database to see if a historical dataset satisfying the calibration parameter range exists, the method further includes: If it does not exist, an initial parameter set that satisfies the calibration parameter range is generated based on a preset sampling algorithm. The initial parameter set includes several sample calibration parameters. Discrete element simulation is performed on each of the sample calibration parameters in the initial parameter set to obtain the corresponding sample simulation physical quantities; The calibration parameters of each sample and the corresponding simulated physical quantities of each sample are stored in the discrete element parameter simulation database and determined as the discrete element simulation dataset.

[0008] In one embodiment, the step of performing discrete element simulation on each of the sample calibration parameters in the initial parameter set to obtain the corresponding sample simulated physical quantity includes: Determine whether there are matching data records for each of the sample calibration parameters in the discrete element parameter simulation database; If so, then determine the sample simulation physical quantity corresponding to the sample calibration parameter that has a matching item based on the matched data record; If not, then perform discrete element simulation on each of the sample calibration parameters to obtain the corresponding sample simulation physical quantities.

[0009] In one embodiment, the step of determining a target proxy model in the proxy model pool based on the calibration parameter space, and predicting the calibration parameter space using the target proxy model to obtain candidate calibration parameters, includes: Matching is performed in the proxy model pool according to the dimension of the calibration parameter space to obtain the target proxy model. Each category of proxy model in the proxy model pool maintains different parameter dimension conditions. The target proxy model is used to perform a global search on the calibration parameter space to obtain the search results; Based on the search results, the calibration parameter that minimizes the difference between the predicted physical quantity and the experimental physical quantity output by the target proxy model is determined in the calibration parameter space as the candidate calibration parameter, wherein the experimental physical quantity is determined based on the current calibration requirements.

[0010] In one embodiment, the step of executing a preset discrete element simulation iterative optimization algorithm using the candidate calibration parameters as the initial iteration point to obtain the target calibration parameters when the current simulated physical quantity does not meet the current calibration requirements includes: Extract experimental physical quantities and error threshold requirements based on the current calibration requirements; Calculate the current relative error based on the experimental physical quantity and the current simulated physical quantity, and determine whether the current relative error meets the error threshold requirement; When the current relative error does not meet the error threshold requirement, a preset discrete element simulation iterative optimization algorithm is executed with the candidate calibration parameters as the initial point of iteration to obtain the target calibration parameters.

[0011] In one embodiment, the step of executing a preset discrete element simulation iterative optimization algorithm using the candidate calibration parameters as the initial iteration point to obtain the target calibration parameters includes: The candidate calibration parameters are used as the initial point for iteration; In each iteration, discrete element simulation is performed on the calibration parameters to be evaluated corresponding to the current iteration to obtain the corresponding iterative simulation physical quantities. Each of the calibration parameters to be evaluated is obtained by iteratively updating the initial point of the iteration. The iteration error is calculated based on the experimental physical quantity and the iterative simulation physical quantity, and it is determined whether the iteration error meets the error threshold requirement. If the conditions are not met, the calibration parameters to be evaluated are updated based on the iteration error, and the next iteration begins. If the conditions are met, the calibration parameters to be evaluated generated in the current iteration will be used as the target calibration parameters.

[0012] In one embodiment, after the step of determining whether the current relative error meets the error threshold requirement, the method further includes: When the current relative error meets the error threshold requirement, the candidate calibration parameter is determined as the target calibration parameter.

[0013] In addition, to achieve the above objectives, this application also proposes a discrete element parameter calibration device, the device comprising: a memory, a processor, and a discrete element parameter calibration program stored in the memory and executable on the processor, wherein the discrete element parameter calibration program, when executed by the processor, implements the discrete element parameter calibration method as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium storing a discrete element parameter calibration program, which, when executed by a processor, implements the discrete element parameter calibration method as described above.

[0015] This application discloses a discrete element method for parameter calibration. The method includes: obtaining the current calibration requirements and initializing the calibration parameter space based on the current calibration requirements; determining the target proxy model in the proxy model pool according to the calibration parameter space, and predicting the calibration parameter space through the target proxy model to obtain candidate calibration parameters. The proxy model pool stores several categories of proxy models trained based on a discrete element simulation dataset; performing discrete element simulation on the candidate calibration parameters to obtain the current simulated physical quantity; and when the current simulated physical quantity does not meet the current calibration requirements, executing a preset discrete element simulation iterative optimization algorithm with the candidate calibration parameters as the initial iteration point to obtain the target calibration parameters.

[0016] Because this application can adaptively determine the target proxy model from the proxy model pool based on the characteristics of the calibration parameter space, it provides a flexible model selection mechanism for different calibration scenarios, enhancing the generalization ability and scenario adaptability of the method. Furthermore, combined with an on-demand iterative optimization strategy starting from the surrogate model prediction, it significantly reduces the number of calls to high-cost discrete element simulations while ensuring that the calibration results meet engineering requirements, thereby improving the computational efficiency and resource utilization of the parameter calibration process. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the first embodiment of the discrete element parameter calibration method of this application; Figure 2 This is a flowchart illustrating the second embodiment of the discrete element parameter calibration method of this application; Figure 3 This is a flowchart illustrating the third embodiment of the discrete element parameter calibration method of this application; Figure 4 This is a schematic diagram of the entire process of the discrete element parameter calibration method in this application; Figure 5This is a schematic diagram of the discrete element parameter calibration device of this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] This application provides a method for calibrating discrete element parameters, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the discrete element parameter calibration method of this application. In this embodiment, the method includes steps S10 to S40: Step S10: Obtain the current calibration requirements and initialize the calibration parameter space based on the current calibration requirements.

[0024] It should be noted that the execution subject of the method in this embodiment can be a computing electronic device with data processing, network communication, and program execution capabilities, such as a discrete element parameter calibration system, a mainframe computer, or a server. It can also be other electronic devices that can be connected to the discrete element parameter calibration system, such as laptops or tablets. This embodiment uses a discrete element parameter calibration device (referred to as the "system") connected to the discrete element parameter calibration system as an example to illustrate the various embodiments of this application.

[0025] It should be understood that the current calibration requirement can be the specific target task for the parameter calibration to be performed, which may include experimental physical quantities, parameter ranges, error thresholds, and fixed parameters.

[0026] Among them, the experimental physical quantity is the target value to be calibrated, such as the static angle of repose of 21.8 degrees; the parameter range is the range of values ​​for each parameter to be calibrated, such as the coefficient of sliding friction between particles. [0.1, 0.8], rolling friction coefficient [0.001, 0.5]; The error threshold is the requirement for calibration accuracy, for example, it can be set to require a relative error of no more than 2%; the fixed parameters can be particle and base plate density, Young's modulus, Poisson's ratio, friction coefficient between particles and base plate, etc.

[0027] It is understandable that the calibration parameter space can be a set of parameters to be calibrated that satisfy the above parameter range. The dimension of the calibration parameter space can be determined according to the categories of the parameters to be calibrated. For example, if the parameters to be calibrated contain two categories of parameter items, it is represented as ( If the calibration parameter space is two-dimensional, then the calibration parameter space can be defined as described above. [0.1, 0.8] represents the horizontal and vertical axes. The two-dimensional plane formed by [0.001, 0.5], in which each point corresponds to a parameter to be calibrated.

[0028] Furthermore, if the parameters to be calibrated also include parameters such as Poisson's ratio, static friction coefficient, and rolling friction coefficient, the dimension of the calibration parameter space can be increased accordingly. This embodiment does not impose any restrictions on this.

[0029] It should also be noted that the system can connect to an SQLite database (hereinafter referred to as the "database"), which can store several calibration configuration files. Each calibration configuration file can be used to characterize different calibration requirements of the user. That is, the user can store different parameter calibration tasks in the database in the form of calibration configuration files, so that the user can directly trigger the execution of a parameter calibration task at any time.

[0030] In its implementation, the system can respond to the user's trigger command for parameter calibration tasks, load and parse the corresponding calibration configuration file from the database, determine the parameter range within it, and then generate a calibration parameter space.

[0031] Step S20: Determine the target proxy model in the proxy model pool according to the calibration parameter space, and predict the calibration parameter space through the target proxy model to obtain candidate calibration parameters. The proxy model pool stores several categories of proxy models trained based on the discrete element simulation dataset.

[0032] It should be noted that the surrogate model pool can be a collection that stores various types of surrogate models, which may include: Kriging Gaussian process regression model, neural network model, support vector machine model, and multinomial response surface model, etc.

[0033] It should also be noted that the various surrogate models can be trained on a discrete element simulation dataset and used to predict physical quantities based on the input calibration parameters. The discrete element simulation dataset can be constructed based on historical data from the aforementioned SQLite database, containing several sample calibration parameters and corresponding sample simulated physical quantities.

[0034] It should be understood that, based on the dimensionality or quantity characteristics of the calibration parameter space, a target proxy model suitable for this parameter calibration task can be selected from the different categories of proxy models mentioned above. Furthermore, the system can use the target proxy model to quickly search and predict the calibration parameter space, obtaining the parameters to be calibrated and their corresponding predicted physical quantities that best approximate the experimental target values.

[0035] Furthermore, to specifically illustrate how to match the target proxy model in the proxy model pool and determine the candidate calibration parameters based on it, step S20 specifically includes: steps S201~S203: Step S201: Match the proxy model in the proxy model pool according to the dimension of the calibration parameter space to obtain the target proxy model. The proxy models of each category in the proxy model pool maintain different parameter dimension conditions.

[0036] It should be noted that the dimension of the calibration parameter space is the number of categories of parameters to be calibrated. For example, if the parameters to be calibrated are the static friction coefficient and the rolling friction coefficient, then the dimension of the calibration parameter space is 2; if the calibration parameters include the elastic modulus, Poisson's ratio, static friction coefficient, and rolling friction coefficient, then the dimension of the calibration parameter space is 4.

[0037] It should be understood that the parameter dimension conditions for each category of surrogate models in the surrogate model pool can be pre-set. For example, the parameter dimension conditions for the Kriging Gaussian process regression model can be set to: low dimension, 2-3 parameters; and the parameter dimension conditions for the neural network model can be set to: high dimension, 4 or more parameters.

[0038] In practice, the dimensions of the calibration parameter space and the parameter dimension conditions of each type of proxy model can be compared, and then at least one proxy model that is suitable for the dimension of the calibration parameter space in this parameter calibration task can be determined from the proxy model pool as the target proxy model.

[0039] Step S202: Perform a global search on the calibration parameter space using the target proxy model to obtain the search results.

[0040] It should be noted that the system can use this target surrogate model to quickly search and predict within the calibration parameter space to obtain search results. Because the surrogate model has low computational cost (milliseconds), it can perform intensive sampling across the entire calibration parameter space or use global optimization algorithms (such as genetic algorithms) to find the optimal region.

[0041] It should be understood that the above search results are a series of intermediate data obtained after the global search is completed, which may include the predicted values ​​of multiple candidate points and their corresponding calibration parameters. If the number of target proxy models is not unique, the target proxy model with the highest weight can be pre-set, and the candidate points (calibration parameters) output by each target proxy model can be deduplicated. The predicted value of the candidate point by the target proxy model with the highest weight can be selected as the predicted physical quantity corresponding to the calibration parameter.

[0042] In the specific implementation, the system can perform a global search of the calibration parameter space through one or more target proxy models, and perform deduplication on the candidate points output by each target proxy model to obtain the predicted physical quantities and corresponding calibration parameters of one or more candidate points after deduplication.

[0043] Step S203: Based on the search results, determine the calibration parameter in the calibration parameter space that minimizes the difference between the predicted physical quantity output by the target proxy model and the experimental physical quantity as the candidate calibration parameter, wherein the experimental physical quantity is determined based on the current calibration requirements.

[0044] It should be noted that for multiple candidate points in the search results above, their corresponding predicted physical quantities can be compared with the actual experimental values ​​(i.e., experimental physical quantities) input by the user in the current calibration requirements. Thus, the predicted physical quantity with the smallest difference from the experimental physical quantity can be determined from the search results, and the calibration parameter corresponding to the predicted physical quantity can be used as the candidate calibration parameter.

[0045] For example, if the selected target surrogate model is the Kriging model, the experimental physical quantity is the experimental angle of repose = 21.8 degrees, and the predicted physical quantity with the smallest difference from the experimental physical quantity is the predicted angle of repose = 21.8; then the calibration parameter corresponding to this predicted physical quantity ( These were identified as candidate calibration parameters.

[0046] Step S30: Perform discrete element simulation on the candidate calibration parameters to obtain the current simulated physical quantity.

[0047] It should be understood that discrete element (DEM) simulation can be achieved based on the DEM simulation software built into the system.

[0048] In practical implementation, the system can use the aforementioned candidate calibration parameters as input to the DEM software, which will then perform DEM simulation to obtain the corresponding current simulated physical quantities.

[0049] For example, the aforementioned =0.2998、 Inputting 0.1129 and other fixed parameters into the DEM software, the current simulated physical quantity can be the static repose angle = 23.66 degrees.

[0050] Step S40: When the current simulated physical quantity does not meet the current calibration requirements, a preset discrete element simulation iterative optimization algorithm is executed with the candidate calibration parameters as the initial point of iteration to obtain the target calibration parameters.

[0051] It should be understood that the system can calculate the relative error between the current simulated physical quantity and the aforementioned experimental physical quantity. If the relative error exceeds the error threshold requirement in the current calibration, it can be considered that the accuracy of the candidate calibration parameter is insufficient, triggering the second stage of optimization.

[0052] It should be noted that the preset discrete element simulation iterative optimization algorithm can be an optimization algorithm based on DEM simulation, such as the least squares method.

[0053] In a specific implementation, the system can take the aforementioned candidate calibration parameters as the starting point, set the maximum number of iterations and the difference step size, and use the least squares method to directly call the DEM simulation to calculate the cost function with relative error as the core. When the maximum number of iterations is reached or the calibration parameters that meet the error threshold requirements are obtained, the target calibration parameters can be determined. The relative error between the simulated physical quantity corresponding to the target calibration parameters and the aforementioned experimental physical quantity meets the error threshold requirements.

[0054] This embodiment can adaptively determine the target surrogate model from the surrogate model pool based on the characteristics of the calibration parameter space, avoiding the problem of insufficient prediction accuracy of a single fixed model under different parameter dimensions or data scales, and improving the initial prediction quality of candidate calibration parameters. At the same time, iterative optimization is only initiated when the prediction results of the surrogate model do not meet the requirements, and the optimization process uses the high-quality candidate points output by the surrogate model as the initial values, reducing unguided blind search and effectively balancing calibration efficiency and accuracy.

[0055] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the discrete element parameter calibration method of this application.

[0056] In this embodiment, to specifically illustrate how to construct a proxy model that meets the current calibration requirements, before step S20, the following steps are included: S01~S03: Step S01: Query the discrete element parameter simulation database to see if there is a historical dataset that meets the calibration parameter range.

[0057] It should be noted that the discrete element parameter simulation database can be the SQLite database mentioned above. This database can also store historical DEM simulation data records or historical datasets created based on historical DEM simulation data records. The historical datasets include parameter-result pairs (calibration parameters and corresponding physical quantities) accumulated from previous calibration tasks.

[0058] It should be understood that the system can first traverse the database to determine whether there exists a historical dataset in which all the calibration parameters fall within the range of the aforementioned calibration parameters.

[0059] For example, in the calibration parameter range of When the range is [0.1, 0.8], if all calibration parameters in a historical dataset in the database are within the above range, then it can be determined that there exists a historical dataset that satisfies the range of calibration parameters.

[0060] Step S02: If it exists, the historical dataset that satisfies the calibration parameter range is determined as the discrete element simulation dataset. The discrete element simulation dataset includes several sample calibration parameters and corresponding sample simulation physical quantities.

[0061] It should be understood that if a historical dataset that meets the calibration parameter range is found, it can be directly loaded from the database as a discrete element simulation dataset for subsequent training of the surrogate model.

[0062] Step S03: Train the initial agent models of different categories using the discrete element simulation dataset, and construct the agent model pool based on the training results.

[0063] It should be noted that the discrete-element simulation dataset used for model training includes several sample pairs, each consisting of sample calibration parameters and the corresponding sample simulated physical quantity. For example, each sample pair can be [( (Static angle of repose).

[0064] It can be understood that different categories of initial surrogate models can be untrained Kriging, neural networks, support vector machines, etc. Specifically, the category selection can be based on the dimensionality of the calibration parameter space in the current parameter calibration task and historical parameter calibration tasks; this embodiment does not impose any restrictions on this. For example, if the dimensions in the parameter calibration task are all low-dimensional, such as one-dimensional or two-dimensional, then only Kriging, which is suitable for low-dimensional input prediction, can be selected as the initial surrogate model.

[0065] In practical implementation, the system can use ( Using the input features and the static rest angle (the specific value of the static rest angle) as the output features, the initial proxy models of different categories are trained to obtain proxy models with predictive capabilities. Then, the trained proxy models can be centrally stored and managed to form a proxy model pool that can be called on demand later.

[0066] Furthermore, if no historical dataset satisfying the aforementioned calibration parameter range exists in the database, then in order to construct a discrete simulation dataset for model training, after step S01, the following steps are also included: steps A021~A023: Step A021: If it does not exist, then generate an initial parameter set that satisfies the calibration parameter range based on the preset sampling algorithm.

[0067] It should be noted that if there is no historical dataset in the database that meets the aforementioned calibration parameter range, the system can directly use a preset sampling algorithm to generate an initial parameter set within the aforementioned calibration parameter range.

[0068] The preset sampling algorithm can be, for example, Latin hypercube or Monte Carlo sampling methods.

[0069] It should be understood that the initial parameter set may include several sample calibration parameters and their associated sample fixed parameters, all of which are within the range of the aforementioned calibration parameters.

[0070] For example, Latin hypercube can be used to generate ten sets of sample calibration parameters within the aforementioned calibration parameter range as an initial dataset to ensure that the initial dataset basically covers the calibration parameter space.

[0071] Step A022: Perform discrete element simulation on each of the sample calibration parameters in the initial parameter set to obtain the corresponding sample simulation physical quantities.

[0072] It should be noted that the system can input the calibration parameters of each sample in the initial parameter set along with the corresponding fixed parameters into the DEM software, thereby performing a complete DEM simulation on each sample calibration parameter and obtaining the corresponding sample simulation physical quantity.

[0073] Furthermore, to prevent duplicate calculations that may occur in concurrent tasks or multiple rounds of operation, before calling the DEM simulation, it is also possible to check again whether the calibration parameters of each sample in the initial parameter set have corresponding data records in the database. Therefore, step A022 also includes: steps A0221~A0223: Step A0221: Determine whether there are matching data records for each of the sample calibration parameters in the discrete element parameter simulation database.

[0074] It should be noted that after generating the initial parameter set, the system can iterate through the data records in the database again to determine if a matching data record exists. This matching data record corresponds to a data record that is completely identical to the sample calibration parameters in the initial parameter set or to calibration parameters that are within the tolerance range.

[0075] Step A0222: If yes, then determine the sample simulation physical quantity corresponding to the sample calibration parameter that has a matching item based on the matched data record.

[0076] It should be understood that if a data record containing a certain sample calibration parameter exists in the database, the corresponding physical quantity can be directly extracted from that data record as the sample simulation physical quantity corresponding to the sample calibration parameter, thereby skipping the subsequent DEM simulation to save system resources.

[0077] Step A0223: If not, perform discrete element simulation on each of the sample calibration parameters to obtain the corresponding sample simulation physical quantities.

[0078] It should be understood that if there are no data records in the database containing any sample calibration parameters, DEM simulation can be performed directly on each sample calibration parameter, and the corresponding physical quantities can be output by the DEM software as the sample simulation physical quantities.

[0079] Step A023: Store the calibration parameters of each sample and the corresponding simulated physical quantities of each sample into the discrete element parameter simulation database and determine it as the discrete element simulation dataset.

[0080] It should be understood that the system can package each sample calibration parameter in the initial dataset and its corresponding sample simulated physical quantity into a sample pair to construct a discrete element simulation dataset for model training.

[0081] Additionally, the sample pairs generated by the above packaging can be stored as a new dataset in the aforementioned database for subsequent use.

[0082] This embodiment queries a discrete element parameter simulation database for historical datasets that meet the calibration parameter range. If such datasets exist, they are directly reused as the discrete element simulation dataset, avoiding redundant sampling and computation. If not, an initial parameter set is generated based on a preset sampling algorithm, and discrete element simulation is performed. The parameter-result pairs are then stored in the database, enabling data construction and expansion. Because this embodiment can perform database matching detection on the sample calibration parameters and directly call historical results for matching items, simulation is only performed on non-repeating parameters, eliminating redundant computation of the same parameter combination. Training different categories of initial proxy models based on the same dataset and constructing a proxy model pool provides a model reserve for subsequent adaptive selection of target proxy models, improving data utilization and model construction efficiency.

[0083] Based on the first and second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to that in embodiments one and two above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating the third embodiment of the discrete element parameter calibration method of this application.

[0084] In this embodiment, to specifically illustrate how to obtain the target calibration parameters through iterative optimization based on candidate calibration parameters, step S40 specifically includes: steps S401~S403: Step S401: Extract the experimental physical quantities and error threshold requirements according to the current calibration requirements.

[0085] It should be noted that the system can extract experimental physical quantities and error threshold requirements from the current calibration requirements. The experimental physical quantities can be values ​​measured in actual experiments, such as a measured static angle of repose of 21.8 degrees. The error threshold requirements are the range of errors allowed by the user, such as relative error. 2%, or absolute error 1 degree.

[0086] Step S402: Calculate the current relative error based on the experimental physical quantity and the current simulated physical quantity, and determine whether the current relative error meets the error threshold requirement.

[0087] It should be understood that the current simulated physical quantity is the actual output obtained after performing DEM simulation on the candidate calibration parameters, such as "the simulated static angle of repose is 23.66 degrees".

[0088] Specifically, the current relative error can be calculated based on the experimental physical quantity and the current simulated physical quantity. The current relative error can be expressed as: relative error = |simulated physical quantity - experimental physical quantity| / experimental physical quantity.

[0089] Therefore, based on the aforementioned example, the current relative error can be calculated as: |23.66-21.8| / 21.8=8.5%.

[0090] In addition, the absolute error can be calculated based on the above example, and the current absolute error can be calculated as: 23.66-21.8=1.86.

[0091] In a practical implementation, the system can compare the calculated current error and / or absolute error with the relative error threshold and / or absolute error threshold in the aforementioned error threshold requirements.

[0092] Step S403: When the current relative error does not meet the error threshold requirement, a preset discrete element simulation iterative optimization algorithm is executed with the candidate calibration parameters as the initial point of iteration to obtain the target calibration parameters.

[0093] It should be understood that if both the current relative error and the current absolute error are less than the aforementioned relative error threshold and / or absolute error threshold, then the candidate calibration parameter can be considered to meet the error threshold requirements. In this case, the aforementioned candidate calibration parameter can be directly used as the target calibration parameter and output.

[0094] If either the current relative error or the current absolute error is less than the aforementioned relative error threshold or absolute error threshold, then the candidate calibration parameters can be considered as not meeting the error threshold requirements. In this case, the second stage of optimization can be triggered, which involves iterative optimization using the candidate calibration parameters as the initial iteration point, iterating until calibration parameters that meet the error threshold requirements are obtained.

[0095] Furthermore, to illustrate in detail how to perform DEM simulation iterative optimization, step S403 specifically includes: steps S4031~S4035: Step S4031: Use the candidate calibration parameters as the initial point for iteration.

[0096] It should be noted that the initial point of this iteration is the starting position of the preset DEM iterative optimization algorithm, thereby improving the search for the optimal target calibration. For example, candidate calibration parameters can be used. =0.2998、 =0.1129 was used as the starting point for the iteration.

[0097] It should be understood that this pre-defined DEM iterative optimization algorithm can be implemented using an optimizer built on the least squares method, which can construct an objective function (cost function) based on the relative error. This objective function can be expressed as:

[0098] in, For each iteration round, simulate the physical quantity. This refers to the physical quantity used in the experiment.

[0099] In each iteration, the system needs to call the DEM simulation once to calculate the physical quantities under the newly generated calibration parameters to be evaluated, and then calculate the error value and send it back to the optimizer for parameter update.

[0100] It should also be noted that an iteration termination condition can be set for the above-mentioned preset DEM iterative optimization algorithm. This iteration termination condition can be set from the following dimensions, including: maximum number of iterations, function tolerance. Variable tolerance and gradient tolerance .

[0101] For example, it can be set =0.1、 =0.001、 The convergence accuracy is set to 0.01, and the maximum number of iterations is set to 50.

[0102] in, =0.1 is used to indicate that if the change in error value is less than 0.1% in two consecutive iterations, it means that the accuracy improvement brought about by iterative optimization has reached its extreme value, and the iteration can be stopped. =0.001 is used to indicate that if the difference between the calibration parameter to be evaluated in the current round and the calibration parameter to be evaluated in the previous round is less than 0.001, it means that the calibration parameter has converged to a stable value and the iteration can be stopped. =0.01 is used to indicate that if the gradient is close to 0.01, it means that the iteration has entered a local optimum and the iteration can be stopped.

[0103] Furthermore, the parameter update method in each iteration can be set, including: the difference step size and the stability loss function. The difference step size can be the direction of the change gradient for generating the calibration parameters to be evaluated in the next iteration. This difference step size can be calculated by solving the Jacobian matrix using the two-point method; for example, the calculated difference step size is 0.1. The stability loss function can be an error calculation method that is not very sensitive to outliers. It can be used to prevent the overall optimization direction from being misled when the physical quantities obtained from a DEM simulation have a large deviation due to numerical noise, thereby ensuring the stability of parameter updates.

[0104] Step S4032: In each iteration, discrete element simulation is performed on the calibration parameters to be evaluated corresponding to the current iteration to obtain the corresponding iterative simulation physical quantities. Each of the calibration parameters to be evaluated is obtained by iteratively updating from the initial point of the iteration.

[0105] It should be noted that in each iteration, the calibration parameters to be evaluated are generated by adjusting the calibration parameters from the previous iteration according to the differential step size. In the initial iteration, the calibration parameters to be evaluated for the current iteration are the aforementioned candidate calibration parameters.

[0106] It should also be noted that before performing DEM simulation, a repeatability check can be performed on the calibration parameter to be evaluated corresponding to the current iteration to determine whether there is a data record in the database containing this calibration parameter. If it is not contained, DEM simulation is performed, and the obtained iterative simulation quantity along with the calibration parameter to be evaluated is written to the database. If it is contained, the corresponding physical quantity in the data record can be directly extracted, and the process jumps to step S4033 to determine whether the error threshold requirement is met.

[0107] In practice, in each iteration, the corresponding calibration parameters to be evaluated, along with the fixed parameters set in the current calibration requirements, can be input into the DEM software to perform DEM simulation, thereby obtaining the iterative simulation physical quantities corresponding to this iteration, such as the static angle of repose.

[0108] Step S4033: Calculate the iteration error based on the experimental physical quantity and the iterative simulation physical quantity, and determine whether the iteration error meets the error threshold requirement.

[0109] In the specific implementation, in each iteration, when the corresponding iterative simulated physical quantity is obtained, its relative error and / or absolute error with the experimental physical quantity are calculated, and then it is determined whether the error threshold requirement is met.

[0110] Step S4034: If the condition is not met, update the calibration parameters to be evaluated based on the iteration error and proceed to the next iteration.

[0111] It should be noted that if the calibration parameters to be evaluated in the current iteration do not meet the aforementioned error threshold requirements, the calibration parameters to be evaluated in the current iteration can be updated based on the aforementioned parameter update method to obtain the calibration parameters to be evaluated in the next iteration and enter the next iteration.

[0112] Step S4035: If satisfied, the calibration parameter to be evaluated generated in the current iteration is used as the target calibration parameter.

[0113] It should be noted that the iteration can be stopped when the calibration parameter to be evaluated in the current iteration meets the error threshold requirement, or when the current iteration meets the aforementioned iteration termination condition. The calibration parameter to be evaluated in the current iteration is then output as the target calibration parameter, and the iterative simulation physical quantity and iteration error corresponding to the target calibration parameter are output simultaneously.

[0114] For example, based on the above iteration, the output target calibration parameters can be: =0.2942、 =0.1026, corresponding to the simulated physical quantity of static repose angle = 22.16 degrees, with an error of 1.65% compared to the experimental physical quantity (static repose angle = 21.8 degrees).

[0115] This embodiment extracts experimental physical quantities and error threshold requirements, calculates the relative error of the current simulated physical quantity and compares it with the threshold, and only initiates iterative optimization when the error does not meet the threshold, avoiding unnecessary calculations for parameters that already meet the accuracy requirements. By using candidate calibration parameters as the initial point of iteration, discrete element simulation is performed on the generated parameters to be evaluated in each iteration and the iteration error is calculated. When the error does not meet the threshold, the parameters are updated according to the optimization algorithm and the iteration continues until the threshold is met, at which point the target calibration parameters are output. This achieves guided local fine search near the surrogate model prediction, improving the parameter accuracy to meet the threshold requirements with as few iterations as possible, and controlling the total number of discrete element simulation calls while ensuring calibration accuracy.

[0116] Furthermore, this can be referenced here. Figure 4 This document provides a complete explanation of the application process. Figure 4 This is a schematic diagram of the entire process of the discrete element parameter calibration method in this application.

[0117] Depend on Figure 4 It can be seen that the discrete element parameter calibration method of this application can be divided into 4 stages: S1: initialization and database preparation; S2: construction of surrogate model; S3: two-stage optimization; S4: result verification and output.

[0118] In S1, the system first initializes in response to the user's trigger command for the parameter calibration task and establishes a synchronous connection with the SQLite database; then it reads the calibration configuration file to load and parse the corresponding calibration configuration file, thereby triggering the parameter calibration task.

[0119] Next, the system can perform repeatability checks in the database: determine whether there is a historical dataset in the database that meets the calibration parameter range conditions (based on the calibration parameters and corresponding physical quantities accumulated from historical calibration tasks). If the initial dataset does not exist, sampling can be performed within the calibration parameter range to generate an initial dataset. The calibration parameter records in the database are updated based on the initial dataset. DEM simulation is then performed on the calibration parameters of each sample in the initial dataset to obtain the corresponding sample simulated physical quantities, thereby constructing a discrete element simulation dataset. The sample simulated physical quantities are then written to the corresponding calibration parameter records in the database. If the initial dataset exists, a discrete element simulation dataset can be generated directly based on the historical dataset.

[0120] In S2, the aforementioned discrete element simulation dataset can be used to train or update different surrogate models, enabling the surrogate models to quickly predict the corresponding simulated physical quantities based on the input parameters.

[0121] The surrogate model can be any one or more combinations of the Kriging Gaussian process regression model, neural network model, support vector machine model, or multinomial response surface model to achieve rapid mapping of the parameter space and result prediction, and can then be based on the trained surrogate model.

[0122] In S3, the trained surrogate model can be used to quickly search and predict the calibration parameter space, and the candidate calibration parameters can be determined based on the prediction results of the surrogate model.

[0123] Next, DEM simulation can be performed on the candidate calibration parameters to obtain the current simulated physical quantity. Then, the relative error between the current simulated physical quantity and the experimental physical quantity can be calculated, and it can be determined whether the relative error is less than the error threshold.

[0124] If the value is not less than the target value, the system default command or the user's real-time command can be used to determine whether to perform the second stage of optimization: DEM simulation iterative optimization is performed with the candidate calibration parameters as the initial point to obtain the target calibration parameters.

[0125] If it is less than, then the candidate calibration parameter can be directly output as the target calibration parameter.

[0126] In S4, to further ensure the reliability of the output target calibration parameters, other error index calculations (such as mean square error, absolute error, etc.) can be performed on the target calibration parameters to verify whether they meet the corresponding threshold requirements.

[0127] Once the target calibration parameters pass the above verification, the target calibration parameters and the corresponding simulated physical quantities can be packaged and stored in the database for subsequent reuse.

[0128] This application's method can achieve parameter-result accumulation and reuse through database management, avoiding multiple DEM simulations with repetitive parameter combinations and significantly reducing computational resource consumption. It utilizes a surrogate model to quickly filter the parameter space, and the two-stage optimization strategy both accelerates the process using the surrogate model and ensures calibration accuracy through DEM iteration, thus resolving the contradiction in existing methods where "pure surrogate models lack accuracy and pure DEM optimization is inefficient." Furthermore, it is applicable to DEM parameter calibration for various particle systems and can directly serve DEM simulation design and performance optimization of industrial equipment, helping to lower the barrier to engineering applications.

[0129] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the discrete element parameter calibration method of this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0130] This application also provides a discrete element parameter calibration device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the discrete element parameter calibration method in Embodiment 1 above.

[0131] The following is for reference. Figure 5 , Figure 5This is a schematic diagram of the discrete element parameter calibration device of this application. The discrete element parameter calibration device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, personal digital assistants (PDAs), tablet computers (PADs), portable media players (PMPs), etc., as well as fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The discrete element parameter calibration device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0132] like Figure 5 As shown, the discrete element parameter calibration device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the discrete element parameter calibration device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the discrete element parameter calibration device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows discrete element parameter calibration devices with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented alternatively.

[0133] The discrete element parameter calibration device provided in this application, employing the discrete element parameter calibration method described in the above embodiments, can solve the technical problem of discrete element parameter calibration. Compared with the prior art, the beneficial effects of the discrete element parameter calibration device provided in this application are the same as those of the discrete element parameter calibration method provided in the above embodiments, and other technical features in this discrete element parameter calibration device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0134] This application also provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the discrete element parameter calibration method in the above embodiments.

[0135] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0136] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described discrete element parameter calibration method, thereby solving the technical problems of the discrete element parameter calibration method. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the discrete element parameter calibration method provided in the above embodiments, and will not be repeated here.

[0137] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other elements in the process, method, article, or system that includes that element.

[0138] The above embodiment numbers are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. They are only some embodiments of this application and do not limit the scope of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included within the protection scope of this application.

Claims

1. A method for calibrating discrete element parameters, characterized in that, The method includes: Obtain the current calibration requirements and initialize the calibration parameter space based on the current calibration requirements; The target proxy model is determined in the proxy model pool based on the calibration parameter space, and the calibration parameter space is predicted by the target proxy model to obtain candidate calibration parameters. The proxy model pool stores several categories of proxy models trained based on the discrete element simulation dataset. Discrete element simulation is performed on the candidate calibration parameters to obtain the current simulated physical quantities; When the current simulated physical quantity does not meet the current calibration requirements, the candidate calibration parameters are used as the initial point of iteration to execute a preset discrete element simulation iterative optimization algorithm to obtain the target calibration parameters.

2. The method as described in claim 1, characterized in that, The current calibration requirements also include: calibration parameter range; before the step of determining the target proxy model in the proxy model pool based on the calibration parameter space, it also includes: Query the discrete element parameter simulation database to see if there is a historical dataset that meets the calibration parameter range; If it exists, the historical dataset that meets the calibration parameter range is determined as the discrete element simulation dataset, which includes several sample calibration parameters and corresponding sample simulation physical quantities. The initial agent models of different categories are trained using the discrete element simulation dataset, and the agent model pool is constructed based on the training results.

3. The method as described in claim 2, characterized in that, After the step of querying the discrete element parameter simulation database to see if a historical dataset exists that satisfies the calibration parameter range, the method further includes: If it does not exist, an initial parameter set that satisfies the calibration parameter range is generated based on a preset sampling algorithm. The initial parameter set includes several sample calibration parameters. Discrete element simulation is performed on each of the sample calibration parameters in the initial parameter set to obtain the corresponding sample simulation physical quantities; The calibration parameters of each sample and the corresponding simulated physical quantities of each sample are stored in the discrete element parameter simulation database and determined as the discrete element simulation dataset.

4. The method as described in claim 3, characterized in that, The step of performing discrete element simulation on each of the sample calibration parameters in the initial parameter set to obtain the corresponding sample simulated physical quantities includes: Determine whether there are matching data records for each of the sample calibration parameters in the discrete element parameter simulation database; If so, then determine the sample simulation physical quantity corresponding to the sample calibration parameter that has a matching item based on the matched data record; If not, then perform discrete element simulation on each of the sample calibration parameters to obtain the corresponding sample simulation physical quantities.

5. The method as described in claim 1, characterized in that, The step of determining a target proxy model in the proxy model pool based on the calibration parameter space, and predicting the calibration parameter space using the target proxy model to obtain candidate calibration parameters includes: Matching is performed in the proxy model pool according to the dimension of the calibration parameter space to obtain the target proxy model. Each category of proxy model in the proxy model pool maintains different parameter dimension conditions. The target proxy model is used to perform a global search on the calibration parameter space to obtain the search results; Based on the search results, the calibration parameter that minimizes the difference between the predicted physical quantity and the experimental physical quantity output by the target proxy model is determined in the calibration parameter space as the candidate calibration parameter, wherein the experimental physical quantity is determined based on the current calibration requirements.

6. The method as described in claim 1, characterized in that, The step of executing a preset discrete element simulation iterative optimization algorithm with the candidate calibration parameters as the initial iteration point to obtain the target calibration parameters when the current simulated physical quantity does not meet the current calibration requirements includes: Extract experimental physical quantities and error threshold requirements based on the current calibration requirements; Calculate the current relative error based on the experimental physical quantity and the current simulated physical quantity, and determine whether the current relative error meets the error threshold requirement; When the current relative error does not meet the error threshold requirement, a preset discrete element simulation iterative optimization algorithm is executed with the candidate calibration parameters as the initial point of iteration to obtain the target calibration parameters.

7. The method as described in claim 6, characterized in that, The step of executing a preset discrete element simulation iterative optimization algorithm using the candidate calibration parameters as the initial iteration point to obtain the target calibration parameters includes: The candidate calibration parameters are used as the initial point for iteration; In each iteration, discrete element simulation is performed on the calibration parameters to be evaluated corresponding to the current iteration to obtain the corresponding iterative simulation physical quantities. Each of the calibration parameters to be evaluated is obtained by iteratively updating the initial point of the iteration. The iteration error is calculated based on the experimental physical quantity and the iterative simulation physical quantity, and it is determined whether the iteration error meets the error threshold requirement. If the conditions are not met, the calibration parameters to be evaluated are updated based on the iteration error, and the next iteration begins. If the conditions are met, the calibration parameters to be evaluated generated in the current iteration will be used as the target calibration parameters.

8. The method as described in claim 6, characterized in that, After the step of determining whether the current relative error meets the error threshold requirement, the method further includes: When the current relative error meets the error threshold requirement, the candidate calibration parameter is determined as the target calibration parameter.

9. A discrete element parameter calibration device, characterized in that, The device includes: a memory, a processor, and a discrete element parameter calibration program stored in the memory and executable on the processor, wherein the discrete element parameter calibration program, when executed by the processor, implements the discrete element parameter calibration method as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium stores a discrete element parameter calibration program, which, when executed by a processor, implements the discrete element parameter calibration method as described in any one of claims 1 to 8.

Images