Discrete element significant mesoscopic parameter screening method based on multi-model fusion

Abstract

The invention discloses a discrete element significant mesoscopic parameter screening method based on multi-model fusion. According to the method, based on a Plackett-Burman test, a Spearman correlation coefficient and a gradient boosting tree algorithm, the contribution degree, correlation and importance degree of a mesoscopic parameter relative to a response value are analyzed and calculated from the perspective of a linear regression model, a rank correlation coefficient and a decision tree, and the one-sidedness of a common method is made up; the Softmax function is adopted to quantify and fuse results of the three types of models, rationality and comprehensiveness of parameter screening are improved, complexity of subsequent parameter calibration tests can be effectively reduced, and powerful support is provided for rapid and accurate establishment of discrete element models.

Description

Technical Field

[0001] The present application relates to the field of simulation technology, and particularly to a discrete element significant mesoscopic parameter screening method, device, equipment and computer-readable storage medium based on multi-model fusion. Background Art

[0002] When carrying out discrete element simulation based on DEM (Discrete Element Method), it involves numerous mesoscopic parameters, which can be classified into three categories: material intrinsic parameters, contact parameters, and contact model parameters. Material intrinsic parameters include Poisson's ratio and shear modulus, etc.; contact parameters mainly include the coefficient of restitution, static friction coefficient, and rolling friction coefficient between materials, etc.; contact model parameters vary depending on the choice of contact model.

[0003] To make the simulation effect approximate to the real physical test, the mesoscopic parameter values need to determine the optimal values. However, too many mesoscopic parameters will cause the number of subsequent parameter calibration tests to increase exponentially, greatly increasing the workload. Therefore, in existing research, for the determination of the optimal values of mesoscopic parameters, the general method is: taking a certain index as the response value, using the Plackett-Burman test to screen significant mesoscopic parameters, the steepest ascent test to determine the optimal interval of significant mesoscopic parameters, and response surface analysis tests such as the Box-Behnken test to obtain the best value; non-significant mesoscopic parameters take the median according to the interval.

[0004] When there are more parameters, the reasonable selection of significant mesoscopic parameters plays an important role in the complexity of subsequent tests. Currently, researchers generally select significant mesoscopic parameters based on the Plackett-Burman test, and the evaluation method is single, and there are certain deviations in the results. Summary of the Invention

[0005] To solve the above technical problems, the present application provides a discrete element significant mesoscopic parameter screening method, device, equipment and computer-readable storage medium based on multi-model fusion.

[0006] In a first aspect, an embodiment of the present application provides a discrete element significant mesoscopic parameter screening method based on multi-model fusion, and the discrete element significant mesoscopic parameter screening method based on multi-model fusion includes:

[0007] Step 1: Solve the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P value of the model, calculating the P value of mesoscopic parameters, and calculating the contribution degree of mesoscopic parameters relative to the response value;

[0008] Step 2: Solve the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of mesoscopic parameters relative to the response value;

[0009] Step 3: Solve the importance of mesoscopic parameters. The specific process includes calculating the mean squared error of the gradient boosting tree model, optimizing the hyperparameters based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value;

[0010] Step 4: Fuse and screen mesoscopic parameters. The specific process includes optimizing the temperature parameters based on information entropy and solving the significance weight assignment of mesoscopic parameters based on the Softmax function.

[0011] Combined with the first aspect, in one implementation, in the model P-value calculation step, the model P-value is used to test whether the overall linear regression model is significant. Select 0.05 as the significance threshold. If the P-value is less than 0.05, the model is considered significant.

[0012] Combined with the first aspect, in one implementation, in the mesoscopic parameter P-value calculation step, the P-value of the mesoscopic parameter is used to determine whether the mesoscopic parameter has a significant impact on the response value. Select 0.05 as the significance threshold. If the P-value is less than 0.05, the mesoscopic parameter is considered significant, and the contribution degree is used to quantify the proportion of the significant impact of each mesoscopic parameter relative to the response value.

[0013] Combined with the first aspect, in one implementation, in the mesoscopic parameter correlation solving step, the Spearman correlation coefficient is selected as the calculation model, and it is not required that the data follow a normal distribution or have a linear relationship.

[0014] Combined with the first aspect, in one implementation, in the hyperparameter optimization step based on the TPE algorithm, with the minimum mean squared error MSE as the objective function, the TPE algorithm is used to automatically search and optimize the hyperparameter combination of the gradient boosting tree model.

[0015] Combined with the first aspect, in one implementation, in the step of calculating the importance of mesoscopic parameters relative to the response value, the gradient boosting tree model is configured with the hyperparameter combination that minimizes the MSE, and the importance of mesoscopic parameters relative to the response value is calculated through the configured gradient boosting tree model.

[0016] Combined with the first aspect, in one implementation, in the mesoscopic parameter fusion and screening step, with the goal that the information entropy of the probability distribution output by the Softmax function approaches a specific entropy value, the temperature parameter T is optimized. Then, based on the optimal temperature parameter, the contribution degree, the absolute value of the correlation coefficient, and the significance weight assignment corresponding to the importance of the mesoscopic parameter relative to the response value are calculated, added to obtain the mesoscopic parameter fusion weight assignment, and the significant mesoscopic parameters are screened based on the fusion weight assignment.

[0017] In the second aspect, an embodiment of the present application provides a discrete element significant mesoscopic parameter screening device based on multi-model fusion. The discrete element significant mesoscopic parameter screening device based on multi-model fusion includes:

[0018] The first solution module is used for solving the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P value of the model, calculating the P value of mesoscopic parameters, and calculating the contribution degree of mesoscopic parameters relative to the response value;

[0019] The second solution module is used for solving the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of mesoscopic parameters relative to the response value;

[0020] The third solution module is used for solving the importance of mesoscopic parameters. The specific process includes calculating the mean square error of the gradient boosting tree model, optimizing hyperparameters based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value;

[0021] The screening module is used for fusing and screening mesoscopic parameters. The specific process includes optimizing the temperature parameter based on information entropy and solving the significance weight assignment of mesoscopic parameters based on the Softmax function.

[0022] In a third aspect, an embodiment of the present application provides a discrete element significant mesoscopic parameter screening device based on multi-model fusion. The discrete element significant mesoscopic parameter screening device based on multi-model fusion includes a processor, a memory, and a discrete element significant mesoscopic parameter screening program based on multi-model fusion stored on the memory and executable by the processor. When the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed by the processor, the steps of the discrete element significant mesoscopic parameter screening method described in the first aspect are implemented.

[0023] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium. A discrete element significant mesoscopic parameter screening program based on multi-model fusion is stored on the computer-readable storage medium. When the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed by a processor, the steps of the discrete element significant mesoscopic parameter screening method described in the first aspect are implemented.

[0024] The beneficial effects brought by the technical solution provided by the embodiment of the present application include:

[0025] The embodiment of the present application is no longer limited to a single Plackett-Burman test for screening significant mesoscopic parameters. Instead, on this basis, the Spearman correlation coefficient and the gradient boosting tree algorithm are used to analyze the significance of mesoscopic parameters relative to the response value from the perspectives of rank correlation coefficient and decision tree, and the Softmax function is used for significance quantification and fusion. Based on multi-model fusion screening, the one-sidedness of general methods can be solved, and at the same time, the complexity of subsequent parameter calibration tests (such as Box-Behnken tests) can be effectively reduced, the test efficiency can be improved, and the rationality and comprehensiveness of discrete element significant mesoscopic parameter screening can be increased. Description of the Drawings

[0026] Figure 1 This is an example diagram of the Plackett - Burman experimental design scheme and result file in a discrete element significant mesoscopic parameter screening method based on multi - model fusion of the present application;

[0027] Figure 2 This is the calculation flow chart of the Plackett - Burman test in a discrete element significant mesoscopic parameter screening method based on multi - model fusion of the present application;

[0028] Figure 3 This is the hyperparameter optimization flow chart based on the TPE algorithm in a discrete element significant mesoscopic parameter screening method based on multi - model fusion of the present application;

[0029] Figure 4 This is the mesoscopic parameter fusion and screening flow chart in a discrete element significant mesoscopic parameter screening method based on multi - model fusion of the present application;

[0030] Figure 5 This is the schematic diagram of the principle of a discrete element significant mesoscopic parameter screening method based on multi - model fusion of the present application;

[0031] Figure 6 This is the schematic diagram of the functional modules of an embodiment of a discrete element significant mesoscopic parameter screening device based on multi - model fusion of the present application;

[0032] Figure 7 This is the schematic diagram of the hardware structure of a discrete element significant mesoscopic parameter screening device based on multi - model fusion involved in the embodiment scheme of the present application. Detailed Embodiment

[0033] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0034] To make the purpose, technical solution and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0035] In a first aspect, an embodiment of the present application provides a discrete element significant mesoscopic parameter screening method based on multi - model fusion.

[0036] In one embodiment, the discrete element significant mesoscopic parameter screening method based on multi - model fusion includes:

[0037] Step 1. Solve the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P-value of the model, calculating the P-value of mesoscopic parameters, and calculating the contribution degree of mesoscopic parameters relative to the response value;

[0038] 1. Solve the contribution degree of mesoscopic parameters

[0039] 1.1. Construct a linear regression model

[0040] In the Plackett-Burman test, each mesoscopic parameter has two values, namely the high-level value and the low-level value. According to the principle of the Plackett-Burman test, n groups of tests can be designed based on the number k of mesoscopic parameters. This part can be completed by self-programming or obtained using software. Refer to Figure 1 , Figure 1 This is an example diagram of the Plackett-Burman test design scheme and result file in a discrete element significant mesoscopic parameter screening method based on multi-model fusion of this application. As Figure 1 shown, assuming that there is a linear relationship between the response value y and k mesoscopic parameters x1, x2,..., x k , it can be expressed as:

[0041] y = β0 + β1x1 + β2x2 +... + β k x k + ε(1)

[0042] In the formula: β0 is the intercept term, β i (i = 1, 2,..., k) is the regression coefficient of the i-th mesoscopic parameter, and ε is the random error term. Assume

[0043] 1.2. Calculate the P-value of the model

[0044] The P-value of model significance is calculated based on the F-test. The F-test is used to compare the explanatory power of the model with the magnitude of random error. In the linear regression model constructed by the Plackett-Burman test, the null hypothesis H0 of the F-test is that the regression coefficients of all mesoscopic parameters in the model are 0, that is, the model is not significant; the alternative hypothesis H1 is that at least one regression coefficient of the mesoscopic parameter is not 0, that is, the model is significant. Calculate the F statistic and determine its corresponding P-value according to the F distribution. If the P-value is less than the given significance threshold (0.05), then reject the null hypothesis and consider the model significant; otherwise, the model is not significant and the subsequent steps should not be carried out. It is necessary to check whether there are errors in the test data. As Figure 2 shown, Figure 2 This is the calculation flow chart of the Plackett-Burman test in a discrete element significant mesoscopic parameter screening method based on multi-model fusion of this application.

[0045] The calculation formula of the F statistic is as follows:

[0046]

[0047] Where: MS model is the mean square of the model, and its calculation formula is: df model is the degree of freedom of the model, which is equal to the number k of mesoscopic parameters; SS model is the sum of squares of the model; MS error is the mean square of error, and its calculation formula is: df error is the degree of freedom of error, which is equal to the number n of test groups minus the number k of mesoscopic parameters minus 1; SS error is the sum of squares of error. Among them, SS model is obtained through the following formula:

[0048]

[0049] Where: is the predicted value of the response value in the i-th test group; is the average of the measured values of the response value; n is the number of test groups.

[0050] SS error is obtained through the following formula:

[0051]

[0052] Where: y i is the measured value of the response value in the i-th test group; is the predicted value of the response value in the i-th test group; n is the number of test groups.

[0053] After obtaining the F statistic, the model P-value is the probability of being greater than or equal to the calculated F statistic under the F distribution with degrees of freedom (df model , df error ), and its calculation formula is as follows:

[0054] P = 1 - F(F stat ; df model , df error )(5)

[0055] Where: P = 1 - F(F stat ; df model , df error ) is the value of the cumulative distribution function of the F distribution with degrees of freedom (df model , df error ) at F stat (the calculated F statistic).

[0056] 1.3. Calculation of the P value of the mesoscopic parameter

[0057] The P value of the mesoscopic parameter is an index used to determine whether the mesoscopic parameter has a significant effect on the response value. For each mesoscopic parameter x i , we want to test whether it has a significant linear effect on the response value y. To this end, the following null hypothesis and alternative hypothesis are established:

[0058] Null hypothesis D0: β i = 0, which means that the mesoscopic parameter x i has no linear effect on the response value y.

[0059] Alternative hypothesis D1: β i ≠ 0, indicating that the mesoscopic parameter x i has a linear effect on the response value y.

[0060] For each mesoscopic parameter x i , the calculation formula for its t statistic is:

[0061]

[0062] Where: β0 is the intercept term in Equation (1); is the estimated value of the regression coefficient of each mesoscopic parameter; is the standard error of each estimated value of the regression coefficient.

[0063] The calculation formula for the P value of the mesoscopic parameter is:

[0064] P = 2 × (1 - F(|t i |; df error )) (7)

[0065] Where: F(|t i |; df error ) is the value of the cumulative distribution function of the t distribution with degrees of freedom df error at |t i |.

[0066] If the P value of the mesoscopic parameter is less than the pre-set significance threshold (0.05), the null hypothesis D0 is rejected, and it is considered that the alternative hypothesis D1 is more likely to hold, that is, the mesoscopic parameter has a significant effect on the response value; conversely, if the P value of the mesoscopic parameter is greater than the significance level, there is not enough evidence to reject the null hypothesis, and only the null hypothesis can be temporarily accepted, believing that the mesoscopic parameter has no significant effect on the response value.

[0067] 1.4. Calculate the contribution degree of the mesoscopic parameter relative to the response value

[0068] Since the significance of the mesoscopic parameters relative to the response value cannot be intuitively obtained through the mesoscopic parameter P value, the contribution degree is used to quantify the proportion of the significant influence of the mesoscopic parameters on the response value. Theoretically, the sorting results of the contribution degrees among the mesoscopic parameters are the same as those of the P value.

[0069] a. Total sum of squares

[0070] The total sum of squares measures the total variation degree of the response value, and its calculation formula is:

[0071]

[0072] In the formula: n is the number of test groups, y i is the measured value of the response value in the i-th test group, is the average of the measured values of the response value.

[0073] b. Partial regression sum of squares

[0074] The partial regression sum of squares is used to measure the contribution of a certain mesoscopic parameter to the variation of the response value when other mesoscopic parameters are fixed. When calculating the partial regression sum of squares SS j of a certain mesoscopic parameter x j it is usually obtained by comparing the residual sum of squares of the complete model including all mesoscopic parameters and the simplified model without this mesoscopic parameter.

[0075] Let the complete model be M1: Its error sum of squares is SSE1; the simplified model M2 is the model after removing the mesoscopic parameter x j : Its error sum of squares is SSE2. Then the partial regression sum of squares of the mesoscopic parameter x j is: SS j = SSE2 - SSE1.

[0076] c. Contribution degree of mesoscopic parameters

[0077] The contribution degree C j of the mesoscopic parameter x j to the response value is defined as the ratio of the partial regression sum of squares of this mesoscopic parameter to the total sum of squares, expressed as a percentage:

[0078]

[0079] Step 2. Solve the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of the mesoscopic parameter relative to the response value;

[0080] 2. Solve the correlation of mesoscopic parameters

[0081] The Spearman correlation coefficient is the rank correlation coefficient. The correlation coefficient ρ of the k-th mesoscopic parameter relative to the response valuek0 , it can be calculated by comparing the sorting differences of each data value of the mesoscopic parameters and the response values, and it is less sensitive to outliers. The calculation formula is shown in Equation (10).

[0082]

[0083] In the formula: x ki is the rank of the k-th mesoscopic parameter corresponding to the i-th group of tests. The rank is the order value obtained after sorting the data values of the mesoscopic parameters and the response values in ascending order among the n groups of tests; is the mean of the ranks of the k-th mesoscopic parameter; y i is the rank of the response value corresponding to the i-th group of tests; is the mean of the ranks of the response value.

[0084] In actual solution, the connection between the mesoscopic parameters and the response value can be ignored. Then, the absolute value of the correlation coefficient ρ of the k-th mesoscopic parameter relative to the response value k can be calculated by a simplified formula, as shown in Equation (11).

[0085]

[0086] In the formula: d ki is the difference in ranks between the k-th mesoscopic parameter and the response value in the i-th group of tests; n is the total number of test groups.

[0087] Step 3: Solve the importance of mesoscopic parameters. The specific process includes calculating the mean squared error of the gradient boosting tree model, optimizing the hyperparameters based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value;

[0088] 3. Solve the importance of mesoscopic parameters

[0089] 3.1 Calculate the mean squared error of the gradient boosting tree model

[0090] The gradient boosting tree is an ensemble model composed of multiple decision trees. The mean squared error (MSE) is a commonly used indicator to measure the prediction error of the model. The calculation formula of MSE is:

[0091]

[0092] In the formula: n is the total number of test groups; y i is the measured value of the response value in the i-th group of tests; is the predicted value of the response value in the i-th group of tests.

[0093] 3.2 Hyperparameter optimization based on the TPE algorithm

[0094] a. Model division

[0095] Taking the minimum MSE as the objective function, the best hyperparameter combination is obtained based on the Bayesian optimization method. The Tree-structured Parzen Estimator (TPE) algorithm is selected to efficiently select the hyperparameter combination, as Figure 3 shown, Figure 3 which is the hyperparameter optimization flowchart based on the TPE algorithm in a discrete element significant mesoscopic parameter screening method based on multi-model fusion in this application. The TPE algorithm selects the next hyperparameter combination by constructing "good" and "bad" models and maximizing the ratio of the two:

[0096] l(x) is the probability model constructed based on the hyperparameter combination with better performance (i.e., the hyperparameter combination with lower MSE), called the "good" model; g(x) is the probability model constructed based on other hyperparameter combinations, called the "bad" model.

[0097] b. Probability model construction

[0098] By comparing the probability densities of these two models, the TPE algorithm selects the hyperparameter combination that is more likely to belong to the "good" model for the next sampling.

[0099] Among them, the TPE algorithm uses the probability density function of the truncated normal distribution to handle the boundary constraints of hyperparameters. For each hyperparameter combination x:

[0100]

[0101] In the formula: μ is the mean vector of the hyperparameter combination; Σ is the covariance matrix; d is the number of hyperparameters; Φ is the cumulative distribution function of the standard normal distribution, and a and b are the truncated lower and upper bound vectors, which are determined according to the value range of the hyperparameters.

[0102] For the "good" model, its probability density is:

[0103]

[0104] In the formula: γ(N) represents the top γ(N) samples with better performance selected from the N evaluated hyperparameter combinations, and these samples are used to construct the "good" model; x i is the hyperparameter combination corresponding to the i-th "good" sample; Σ i is the covariance matrix of the i-th "good" sample; a i , b i are the truncated lower and upper bound vectors of the i-th "good" sample respectively.

[0105] For the "bad" model, its probability density is:

[0106]

[0107] Where: N - γ(N) represents the number of remaining hyperparameter combinations with poor performance, and these samples are used to construct the "bad" model; x i is the hyperparameter combination corresponding to the i-th "bad" sample; Σ i is the covariance matrix of the i-th "bad" sample; a i and b i are the lower and upper bound vectors of truncation for the i-th "bad" sample, respectively.

[0108] c, sampling strategy

[0109] The criterion for selecting the next hyperparameter combination x is to maximize the ratio of the "good" model to the "bad" model, and this ratio reflects the degree to which the hyperparameter combination x belongs to the "good" model:

[0110]

[0111] Where: X is the value space of hyperparameter combinations.

[0112] After successfully selecting the next hyperparameter combination, calculate the MSE corresponding to this hyperparameter combination until the MSE is basically unchanged, then it is considered that the MSE reaches the minimum value that can be achieved under the current model, and the optimal hyperparameter combination is obtained.

[0113] 3.3. Calculate the importance of mesoscopic parameters relative to the response value

[0114] After obtaining the optimal hyperparameter combination, use it to configure the model. Assume that for the mesoscopic parameter k, in the m-th decision tree, its contribution to the reduction of the loss function in all node splits is I km . Then, in the entire gradient boosting tree model, the importance F Ik of the mesoscopic parameter k relative to the response value can be obtained by accumulating and normalizing its contributions in all decision trees:

[0115]

[0116] Where: M is the total number of trees, and K is the total number of mesoscopic parameters.

[0117] Step Four: Mesoscopic parameter fusion and screening, and the specific process includes temperature parameter optimization based on information entropy and solution of the significance weight assignment of mesoscopic parameters based on the Softmax function.

[0118] 4. Mesoscopic parameter fusion and screening

[0119] To uniformly compare the results of the Plackett - Burman test, Spearman correlation coefficient, and gradient boosting tree algorithm, that is, the contribution degree, correlation coefficient, and importance degree of mesoscopic parameters relative to the response value, the Softmax function is used to assign significant weights to the results of the three types of models, as Figure 4 shown, Figure 4 This is the flow chart of mesoscopic parameter fusion and screening in a discrete element significant mesoscopic parameter screening method based on multi - model fusion in this application. For the result z=(z1, z2, …, z k ) of each model, the Softmax function amplifies the weights of larger values through exponential transformation to generate a probability distribution y=(y1, y2, …, y k ), that is, the significant weight assignment. The formula is:

[0120]

[0121] In the formula: T is the temperature parameter, which is used to control the "sharpness" of the probability distribution. Among them, the results of the three types of models are the contribution degree of mesoscopic parameters relative to the response value, the absolute value of the correlation coefficient, and the importance degree respectively. Therefore, the number of z is the same as the number of mesoscopic parameters.

[0122] To find the appropriate temperature parameter T, the temperature parameter T is optimized with the goal that the information entropy of the probability distribution output by the Softmax function approaches a specific entropy value. For y=(y1, y2, …, y k ), the probabilities corresponding to each value are p1, p2, …, p k , and satisfy The corresponding information entropy H(X) is defined as:

[0123]

[0124] The numerical optimization method is used to continuously and automatically optimize to obtain the optimal temperature parameter. Based on this, the significant weight assignments corresponding to the results of the three types of models are calculated. The corresponding weight assignments of mesoscopic parameters under the three types of models are added together to obtain the mesoscopic parameter fusion and screening results. The mesoscopic parameters with larger weight assignments are significant mesoscopic parameters.

[0125] Since the results (contribution degree, correlation coefficient, and importance degree) under different models are different, the optimal temperature parameter T obtained by the numerical optimization method is also different, resulting in differences in formula (18).

[0126] For the result z=(z1, z2, …, z k) Through Equation (18), this result will be transformed into a significance weight assignment, obtaining a weight assignment ranging from 0 to 1, which is based on the internal differences of this model result.

[0127] Specifically, the results of the three models are all weighted using the Softmax function. Then, the corresponding weight assignments for each mesoscopic parameter (a total of 3 weight assignments) are added together, which is the significance weight of the mesoscopic parameter relative to the response value. Sorting them from largest to smallest gives the screening result of the mesoscopic parameter fusion.

[0128] The judgment criteria for significant mesoscopic parameters are not strictly restricted in the embodiments of this application. It is possible to determine the pre-set number or pre-set proportion of mesoscopic parameters as significant mesoscopic parameters according to the sorting differences of the significance weights of the mesoscopic parameters.

[0129] Refer to Figure 5 , Figure 5 is the schematic diagram of the principle of a method for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion in this application. Combining Figures 1 to 5 As shown, the embodiments of this application construct a method for screening significant mesoscopic parameters of the discrete element method based on Plackett-Burman test, Spearman correlation coefficient, and gradient boosting tree algorithm. They analyze and calculate the contribution degree, correlation, and importance of mesoscopic parameters relative to the response value from the perspectives of linear regression model, rank correlation coefficient, and decision tree respectively, and use the Softmax function to quantify and fuse the results of the three types of models, improving the rationality and comprehensiveness of parameter screening, effectively reducing the complexity of subsequent parameter calibration experiments, and providing strong support for the rapid and accurate establishment of the discrete element model.

[0130] In the second aspect, the embodiments of this application also provide a device for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion.

[0131] In one embodiment, refer to Figure 6 , Figure 6 is the schematic diagram of the functional modules of an embodiment of the device for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion in this application. As Figure 6 shown, the device for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion includes:

[0132] The first solving module 10 is used for solving the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P value of the model, calculating the P value of the mesoscopic parameter, and calculating the contribution degree of the mesoscopic parameter relative to the response value;

[0133] The second solving module 20 is used for solving the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of the mesoscopic parameter relative to the response value;

[0134] The third solution module 30 is used for solving the importance of mesoscopic parameters. The specific process includes calculating the mean square error of the gradient boosting tree model, optimizing the hyperparameters based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value.

[0135] The screening module 40 is used for fusing and screening mesoscopic parameters. The specific process includes optimizing the temperature parameters based on information entropy and solving the significance weight assignment of mesoscopic parameters based on the Softmax function.

[0136] Further, in one embodiment, the P-value of the model is used to test whether the overall linear regression model is significant. A significance threshold of 0.05 is selected. If the P-value is less than 0.05, the model is considered significant.

[0137] Further, in one embodiment, the P-value of the mesoscopic parameter is used to determine whether the mesoscopic parameter has a significant impact on the response value. A significance threshold of 0.05 is selected. If the P-value is less than 0.05, the mesoscopic parameter is considered significant, and the contribution degree is used to quantify the proportion of the significant impact of each mesoscopic parameter relative to the response value.

[0138] Further, in one embodiment, in the step of solving the correlation of mesoscopic parameters, the Spearman correlation coefficient is selected as the calculation model, and it is not required that the data follow a normal distribution or have a linear relationship.

[0139] Further, in one embodiment, in the step of optimizing the hyperparameters based on the TPE algorithm, the minimum mean square error MSE is used as the objective function, and the TPE algorithm is used to automatically search for and optimize the hyperparameter combination of the gradient boosting tree model.

[0140] Further, in one embodiment, in the step of calculating the importance of mesoscopic parameters relative to the response value, the gradient boosting tree model is configured with the hyperparameter combination that minimizes the MSE, and the importance of mesoscopic parameters relative to the response value is calculated through the configured gradient boosting tree model.

[0141] Further, in one embodiment, in the step of fusing and screening mesoscopic parameters, with the goal that the information entropy of the probability distribution output by the Softmax function approaches a specific entropy value, the temperature parameter T is optimized. Then, based on the optimal temperature parameter, the contribution degree, the absolute value of the correlation coefficient, and the significance weight assignment corresponding to the importance of mesoscopic parameters relative to the response value are calculated, added to obtain the mesoscopic parameter fusion weight assignment, and significant mesoscopic parameters are screened based on the fusion weight assignment.

[0142] Among them, the function implementation of each module in the above-mentioned discrete element significant mesoscopic parameter screening device based on multi-model fusion corresponds to each step in the embodiment of the above-mentioned discrete element significant mesoscopic parameter screening method based on multi-model fusion, and its function and implementation process will not be elaborated here one by one.

[0143] In a third aspect, an embodiment of the present application provides a discrete element significant mesoscopic parameter screening device based on multi-model fusion. The discrete element significant mesoscopic parameter screening device based on multi-model fusion can be a device with data processing functions such as a personal computer (PC), a laptop computer, a server, etc.

[0144] Referring to Figure 7 , Figure 7 is a schematic hardware structure diagram of the discrete element significant mesoscopic parameter screening device involved in the solution of the embodiment of the present application. In the embodiment of the present application, the discrete element significant mesoscopic parameter screening device based on multi-model fusion may include a processor, a memory, a communication interface, and a communication bus.

[0145] Among them, the communication bus can be of any type and is used to interconnect the processor, the memory, and the communication interface.

[0146] The communication interface includes interfaces such as input / output (I / O) interfaces, physical interfaces, and logical interfaces for implementing the interconnection of components inside the discrete element significant mesoscopic parameter screening device based on multi-model fusion, and interfaces for implementing the interconnection of the discrete element significant mesoscopic parameter screening device based on multi-model fusion with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, an optical fiber interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.

[0147] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0148] The processor may be a general-purpose processor, which can call the discrete element significant mesoscopic parameter screening program based on multi-model fusion stored in the memory and execute the discrete element significant mesoscopic parameter screening method based on multi-model fusion provided by the embodiments of the present application. For example, the general-purpose processor may be a central processing unit (CPU). Among them, the method executed when the discrete element significant mesoscopic parameter screening program based on multi-model fusion is called may refer to the various embodiments of the discrete element significant mesoscopic parameter screening method based on multi-model fusion in the present application, which will not be elaborated here.

[0149] Those skilled in the art can understand that Figure 7 the hardware structure shown in does not constitute a limitation to the present application, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0150] In a fourth aspect, the embodiments of the present application further provide a computer-readable storage medium.

[0151] The computer-readable storage medium of the present application stores a discrete element significant mesoscopic parameter screening program based on multi-model fusion. When the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed by a processor, the steps of the discrete element significant mesoscopic parameter screening method based on multi-model fusion as described above are implemented.

[0152] Among them, the method implemented when the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed may refer to the various embodiments of the discrete element significant mesoscopic parameter screening method based on multi-model fusion in the present application, which will not be elaborated here.

[0153] It should be noted that the serial numbers of the above embodiments of the present application are only for description and do not represent the superiority or inferiority of the embodiments.

[0154] The terms "including" and "having" and any variations thereof in the specification, claims and drawings of the present application are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include unlisted steps or units, or may optionally further include other steps or units inherent to these processes, methods, products or devices. The descriptions with terms such as "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit that "first", "second" and "third" are of different types.

[0155] In the description of the embodiments of the present application, words such as "exemplary", "for example", or "for illustration" are used to indicate examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary", "for example", or "for illustration" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary", "for example", or "for illustration" is intended to present relevant concepts in a specific manner.

[0156] In the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B. The "and / or" in the text is merely a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B may mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present application, "a plurality of" means two or more than two.

[0157] In some processes described in the embodiments of the present application, a plurality of operations or steps appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or may be executed in parallel. The serial numbers of the operations are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in order or in parallel, and these operations or steps may be combined.

[0158] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) as described above, and includes several instructions for causing a terminal device to execute the methods described in the various embodiments of the present application.

[0159] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made by using the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A screening method for discrete element significant mesoscopic parameters based on multi-model fusion, characterized in that The method for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion includes: Step 1, solving the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P-value of the model, calculating the P-value of mesoscopic parameters, and calculating the contribution degree of mesoscopic parameters relative to the response value; Step 2, solving the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of mesoscopic parameters relative to the response value; Step 3, solving the importance of mesoscopic parameters. The specific process includes calculating the mean square error of the gradient boosting tree model, hyperparameter optimization based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value; Step 4, fusing and screening mesoscopic parameters. The specific process includes optimizing the temperature parameter based on information entropy and solving the significance weight assignment of mesoscopic parameters based on the Softmax function.

2. The method for screening discrete element significant mesoscopic parameters based on multi-model fusion as described in claim 1, wherein In the step of calculating the P-value of the model, the P-value of the model is used to test whether the overall linear regression model is significant. A significance threshold of 0.05 is selected. If the P-value is less than 0.05, the model is considered significant.

3. The method for screening discrete element significant mesoscopic parameters based on multi-model fusion according to claim 1, wherein In the step of calculating the P-value of mesoscopic parameters, the P-value of mesoscopic parameters is used to judge whether the mesoscopic parameter has a significant impact on the response value. A significance threshold of 0.05 is selected. If the P-value is less than 0.05, the mesoscopic parameter is considered significant, and the contribution degree is used to quantify the proportion of the significant impact of each mesoscopic parameter relative to the response value.

4. The method for screening significant mesoscopic parameters of discrete element based on multi-model fusion according to claim 1, characterized in that, In the step of solving the correlation of mesoscopic parameters, the Spearman correlation coefficient is selected as the calculation model, and it is not required that the data follow a normal distribution or have a linear relationship.

5. The method for screening discrete element significant mesoscopic parameters based on multi-model fusion according to claim 1, wherein In the step of hyperparameter optimization based on the TPE algorithm, with the minimum mean square error MSE as the objective function, the TPE algorithm is used to automatically search and optimize the hyperparameter combination of the gradient boosting tree model.

6. The method for screening discrete element significant mesoscopic parameters based on multi-model fusion according to claim 1, wherein In the step of calculating the importance of mesoscopic parameters relative to the response value, the hyperparameter combination that minimizes the MSE is used to configure the gradient boosting tree model, and the importance of mesoscopic parameters relative to the response value is calculated through the configured gradient boosting tree model.

7. The method for screening discrete element significant mesoscopic parameters based on multi-model fusion according to claim 1, wherein In the step of fusing and screening mesoscopic parameters, with the goal that the information entropy of the probability distribution output by the Softmax function approaches a specific entropy value, the temperature parameter T is optimized. Then, based on the optimal temperature parameter, the significance weight assignments corresponding to the contribution degree, the absolute value of the correlation coefficient, and the importance of mesoscopic parameters relative to the response value are calculated, added to obtain the fusion weight assignment of mesoscopic parameters, and significant mesoscopic parameters are screened based on the fusion weight assignment.

8. A discrete element significant mesoscopic parameter screening device based on multi-model fusion, characterized in that, The device for screening significant mesoscopic parameters of the discrete element method based on multi-model fusion includes: The first solving module is used to solve the contribution degree of mesoscopic parameters. The specific process includes constructing a linear regression model, calculating the P-value of the model, calculating the P-value of mesoscopic parameters, and calculating the contribution degree of mesoscopic parameters relative to the response value; The second solving module is used to solve the correlation of mesoscopic parameters. The specific process includes calculating the absolute value of the correlation coefficient of mesoscopic parameters relative to the response value; The third solving module is used to solve the importance of mesoscopic parameters. The specific process includes calculating the mean square error of the gradient boosting tree model, hyperparameter optimization based on the TPE algorithm, and calculating the importance of mesoscopic parameters relative to the response value; A screening module for mesoscopic parameter fusion screening, and the specific process includes temperature parameter optimization based on information entropy and solving the significance weight assignment of mesoscopic parameters based on the Softmax function.

9. A discrete element significant mesoscopic parameter screening device based on multi-model fusion, characterized in that, The discrete element significant mesoscopic parameter screening device based on multi-model fusion includes a processor, a memory, and a discrete element significant mesoscopic parameter screening program based on multi-model fusion stored on the memory and executable by the processor. When the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed by the processor, the steps of the discrete element significant mesoscopic parameter screening method based on multi-model fusion as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that, A discrete element significant mesoscopic parameter screening program based on multi-model fusion is stored on the computer-readable storage medium. When the discrete element significant mesoscopic parameter screening program based on multi-model fusion is executed by the processor, the steps of the discrete element significant mesoscopic parameter screening method based on multi-model fusion as described in any one of claims 1 to 7 are implemented.

Images