Method, device and equipment for predicting compression molding quality of granular mixtures and medium

Abstract

The invention discloses a granular mixture compression molding quality prediction method, device, equipment and medium, and relates to the technical field of energetic material and explosive preparation, the method constructs a prediction model based on the relevance between key process parameters and molded grain density, can predict and analyze the density and density distribution after granular mixture compression molding, and can predict and analyze the quality of granular mixture compression molding. The method can effectively make up for the deficiency of characterization of the characteristic parameters of the granular mixture in the compact non-transparent confined space forming process by the current actual characterization means, and can provide visual and accurate real-time data support for research on the change rule of the temperature field and the stress field of the granular mixture in the compression forming process under different process conditions. And a theoretical foundation can be laid for accurate control of forming process parameters and further improvement of forming quality.

Description

Technical Field

[0001] The invention relates to the technical field of energetic materials and explosive preparation, and in particular to a method, device, equipment and medium for predicting the quality of bulk body compression molding. Background Art

[0002] The compaction process of explosive bulk is to use external pressure to squeeze the bulk in a mold or shell to make it compact, so that it has a certain shape and size, and endow it with the required mechanical properties and other functional characteristics. Under the thermal coupling of compaction, the gap between the bulk particles gradually decreases, the contact area increases, and its accumulation tends to the minimum volume bonding filling. This process is affected by the molding process parameters such as temperature, pressure, and time.

[0003] Therefore, the molding process parameters are the key factors affecting the quality of the charge density. Inaccurate molding process parameters can easily lead to charge quality problems such as many internal defects and low density in the molded granular body.

[0004] Therefore, how to provide a method for predicting the quality of explosive bulk compaction, which can effectively make up for the shortcomings of existing means in characterizing the compaction quality of bulk explosives in dense non-transparent confined spaces, and realize real-time regularity analysis of the charging quality of explosive bulk and effective regulation of the charging quality, is a technical problem that urgently needs to be solved by technical personnel in this field. Summary of the invention

[0005] In view of the above problems, the present invention provides a method, device, equipment and medium for predicting the quality of granular body compaction for overcoming the above problems or at least partially solving the above problems.

[0006] The present invention provides the following scheme:

[0007] A method for predicting the quality of granular body compacting, comprising:

[0008] The principal component analysis method is used to transform the variables with high correlation in the granular pressing process into independent or low-correlation variables to determine the three key parameters required to build a molding quality prediction model.

[0009] The correlation between each of the key parameters and the molding density is determined by using a scatter analysis method;

[0010] The density data of the formed drug column is obtained by combining the three key parameters through a three-factor six-level orthogonal test;

[0011] The abnormal parameter analysis method is used to eliminate the abnormal data generated during the data collection process according to the 3σ principle;

[0012] The maximum - minimum normalization method is used to unify the data dimension without changing the distribution characteristics of the original process parameters to obtain sample data;

[0013] The support vectors and hyperplane are characterized through the solution process of the Lagrangian function, and the fitting error and the distance from the support vectors to the hyperplane are minimized to construct a support vector regression machine model. The support vector regression machine model includes a smoothing function for predicting continuous density values;

[0014] In the regression problem, a loss function is constructed to measure the fitting effect of the support vector regression machine model during the training process. The sample data is brought in for the training of the support vector regression machine model, and the Bayesian optimization method is used to optimize the model performance to obtain the optimal parameters, so as to obtain the forming quality prediction model;

[0015] Test data is input into the forming quality prediction model so that the forming quality prediction model outputs a forming quality prediction result; the test data includes at least three test parameters, and the types and orders of the three test parameters are the same as those of the three key parameters in the sample data.

[0016] Preferably: The principal component analysis method includes calculating the principal component contribution rate of each process parameter to the forming density by the principal component cumulative contribution rate formula, and the contribution rate is calculated by the following formula:

[0017]

[0018] In the formula: Q (p) represents the cumulative contribution rate of the first p principal components, and λ k represents the eigenvalue.

[0019] Preferably: The three key parameters include the forming pressure, the forming temperature, and the holding pressure time.

[0020] Preferably: The determination of the correlation between each key parameter and the forming density by the scatter analysis method includes:

[0021] Normalize each variable to eliminate the dimension difference between variables, fit the regression curve between each key parameter and the forming density, and represent the correlation with the forming density through the slope.

[0022] Preferably: The kernel function of the support vector regression machine model is a Gaussian kernel function.

[0023] Preferably: The support vector regression machine model is represented by the following formula:

[0024]

[0025] In the formula: a i 、 is the Lagrange multiplier, and k(x i , x) represents the kernel function of the support vector regression machine, y j represents the density value in the model, and ε is the tolerance.

[0026] Preferably: The Bayesian optimization method is represented by the following formula:

[0027]

[0028] In the formula: A and B respectively represent two random events, P(A i |B) represents the conditional probability of A occurring after the known event B occurs, that is, P(A i |B) is the posterior probability of A, and P(B|A i ) is the likelihood function of B; P(A i ) and P(A i ) are the prior probabilities of A without considering the influence of event B. i ) and P(A j )

[0029] A granular material compression molding quality prediction device for performing the above-mentioned granular material compression molding quality prediction method, the device includes:

[0030] A key parameter determination unit for using the principal component analysis method to transform the variables with high correlation in the granular material compression molding process into variables that are independent of each other or have low correlation, so as to determine three key parameters required for constructing a molding quality prediction model;

[0031] A correlation calculation unit for using the scatter plot analysis method to determine the correlation between each of the key parameters and the molding density;

[0032] A density data acquisition unit for obtaining the density data of the molded charge by means of a three-factor six-level orthogonal experiment in combination with the three key parameters;

[0033] An abnormal data rejection unit for using the abnormal parameter analysis method to reject the abnormal data generated during the data acquisition process according to the 3σ principle;

[0034] A sample data acquisition unit for using the maximum-minimum normalization method to unify the data dimension without changing the distribution characteristics of the original process parameters to obtain sample data;

[0035] A model construction unit for depicting the support vector and the hyperplane through the solution process of the Lagrangian function, and minimizing the fitting error and the distance from the support vector to the hyperplane, so as to construct a support vector regression machine model, and the support vector regression machine model includes a smooth function for predicting continuous density values;

[0036] A model optimization unit is used to construct a loss function in a regression problem to measure the fitting effect of the support vector regression machine model during training, input the sample data for training the support vector regression machine model, and optimize the model performance using the Bayesian optimization method to obtain the optimal parameters, so as to obtain the forming quality prediction model;

[0037] A quality prediction unit is used to input test data into the forming quality prediction model, so that the forming quality prediction model outputs a forming quality prediction result; the test data includes at least three test parameters, and the types and orders of the three test parameters are the same as those of the three key parameters in the sample data.

[0038] A granular material compression molding quality prediction device, the device includes a processor and a memory:

[0039] The memory is used to store program codes and transmit the program codes to the processor;

[0040] The processor is used to execute the above-mentioned granular material compression molding quality prediction method according to the instructions in the program codes.

[0041] A computer-readable storage medium is used to store program codes, and the program codes are used to execute the above-mentioned granular material compression molding quality prediction method.

[0042] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0043] A granular material compression molding quality prediction method, device, equipment and medium provided by an embodiment of the present application. This method constructs a prediction model based on the correlation between key process parameters and the density of the formed charge, and can predict and analyze the density and density distribution after the granular material is compressed and molded. It can effectively make up for the deficiency of the current actual characterization means for characterizing the characteristic parameters of the granular material during the forming process in a dense and non-transparent confined space, and can provide intuitive and accurate real-time data support for the research on the change laws of the temperature field and stress field of the granular material during the compression molding process under different process conditions, and lay a theoretical foundation for the precise control of the forming process parameters and the further improvement of the forming quality.

[0044] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages at the same time. Description of the Drawings

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 It is a flow chart of a method for predicting the quality of granular body pressing and forming provided by an embodiment of the present invention;

[0047] Figure 2 It is a flow chart of parameter optimization of support vector regression machine based on Bayesian optimization provided by an embodiment of the present invention;

[0048] Figure 3 It is a thermal diagram of the correlation between key process parameters and molding density provided by an embodiment of the present invention;

[0049] Figure 4 is a schematic diagram of a model training hyperplane provided in an embodiment of the present invention;

[0050] Figure 5 is a schematic diagram of a hyperplane of a model with a holding time of 35 minutes provided in an embodiment of the present invention;

[0051] Figure 6 is a schematic diagram of a granular body compression molding quality prediction device provided by an embodiment of the present invention;

[0052] Figure 7 It is a schematic diagram of a bulk body pressing molding quality prediction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0054] See also Figure 1 , is a method for predicting the quality of granular body pressing provided by an embodiment of the present invention, such as Figure 1 As shown, the method may include:

[0055] S101: The variables with high correlation in the granular body pressing process are transformed into independent or low-correlated variables by using the principal component analysis method to determine the three key parameters required for constructing the molding quality prediction model; in specific implementation, the embodiment of the present application can provide that the principal component analysis method includes calculating the principal component cumulative contribution rate formula to obtain the principal component contribution rate of each process parameter to the molding density, and the contribution rate is calculated by the following formula:

[0056]

[0057] Where: Q (p) represents the cumulative contribution rate of the first p principal components, λ k Represents the eigenvalue.

[0058] According to actual calculations, the three key parameters include molding pressure, molding temperature and holding time.

[0059] S102: Use scatter analysis method to determine the correlation between each of the key parameters and the molding density; in specific implementation, the embodiment of the present application can provide normalization of each variable, eliminate the dimensional difference between each variable, fit the regression curve between each of the key parameters and the molding density, and express the correlation with the molding density through the slope.

[0060] S103: combining the three key parameters and performing a three-factor six-level orthogonal test to obtain density data of the formed drug column;

[0061] S104: using an abnormal parameter analysis method to eliminate abnormal data generated during the data collection process according to the 3σ principle;

[0062] S105: Using the maximum and minimum normalization method to unify the data dimension without changing the original process parameter distribution characteristics to obtain sample data;

[0063] S106: Characterize the support vector and the hyperplane through the Lagrangian function solution process, and minimize the fitting error and the distance from the support vector to the hyperplane to construct a support vector regression machine model, wherein the support vector regression machine model includes a smoothing function for predicting continuous density values; in specific implementation, the embodiment of the present application can provide that the kernel function of the support vector regression machine model is a Gaussian kernel function.

[0064] The support vector regression model is represented by the following formula:

[0065]

[0066] Where: a i , is the Lagrange multiplier, k(x i , x) represents the kernel function of the support vector regression machine, y j Represents the density value in the model, and ε is the tolerance.

[0067] S107: constructing a loss function in the regression problem to measure the fitting effect of the support vector regression machine model during the training process, bringing in the sample data to train the support vector regression machine model, and using the Bayesian optimization method to optimize the model performance to obtain the optimal parameters so as to obtain the molding quality prediction model; in specific implementation, the embodiment of the present application can provide that the Bayesian optimization method is expressed by the following formula:

[0068]

[0069] Where: A and B represent two random events, P(A i |B) indicates that after the known event B occurs, A i The conditional probability of occurrence, that is, P(A i |B) is A i The posterior probability, P(B|A i ) is the likelihood function of B; P(A i ) and P(A j ) does not consider the impact of event B on the prior probability of A.

[0070] S108: Inputting test data into the molding quality prediction model so that the molding quality prediction model outputs a molding quality prediction result; the test data includes at least three test parameters, and the types and order of the three test parameters are the same as those of the three key parameters in the sample data.

[0071] The method for predicting the quality of granular compression molding provided in the embodiment of the present application adopts the principal component analysis method to determine the key parameters of the compression molding process, adopts the scatter analysis method to determine the correlation between each key parameter and the molding density, obtains the density data of the molded medicine column through a three-factor six-level orthogonal test, adopts the abnormal parameter analysis method to eliminate the abnormal data generated in the data acquisition process according to the 3σ principle, adopts the maximum and minimum normalization method to unify the data dimension without changing the original process parameter distribution characteristics, adopts the support vector regression model for training, constructs the loss function in the regression problem to measure the fitting effect of the regression model in the training process, introduces the data sample for regression model training, adopts the Bayesian optimization method to optimize the model performance, and obtains the optimal parameters.

[0072] The method provided by this application is described in detail below. Figure 2 shown.

[0073] This method is based on machine learning theory, vector machine theory and algorithms. It collects key parameters such as forming temperature, pressure, and holding time during the pressing process, conducts abnormal parameter analysis and elimination, and normalizes the key parameters. This method is trained by the vector machine regression method, and the model is further optimized by adjusting the kernel function, regularization parameter, and tolerance to improve the accuracy of the forming quality prediction model. When solving the internal parameters of the model, the forming quality prediction model and the optimization method can also iteratively fit the relationship between the key process parameters and the forming density in a cycle, with high prediction accuracy, and can provide a basis for the selection of forming process parameters.

[0074] First, the principal component analysis method is used to transform variables with high correlation into independent or less correlated variables, interpret most of the variables in the original data with fewer variables, achieve multi-variable dimensionality reduction analysis, and determine that the key process parameters affecting the forming density are forming pressure, forming temperature, and holding time.

[0075] Using the principal component analysis method, calculated by the formula of the cumulative contribution rate of the principal components, the contribution rate of each process parameter to the forming density is obtained. The calculation formula for the cumulative contribution rate Q(p) of the first p principal components is:

[0076]

[0077] Among them, Q (p) represents the cumulative contribution rate of the first p principal components, and λ k represents the eigenvalue.

[0078] The calculation results show that the contribution rate of the pressing forming pressure is 68.29%, the forming temperature is 23.23%, and the holding time is 8.49%. According to the principal component analysis method, the components with a cumulative contribution of not less than 95% should be selected as the principal components. Thus, it can be determined that the forming pressure, forming temperature, and holding time are the principal component variables affecting the forming density of granular materials, that is, the key process parameters.

[0079] The scatter analysis method is used to determine the influence degree of each key parameter on the forming density. Compare the regression magnitude between each variable and the forming density, normalize each variable, and eliminate the dimensional difference between variables. The scatter analysis method determines the correlation between key parameters such as forming pressure, forming temperature, and holding time and the forming density, normalizes each variable, eliminates the dimensional difference between variables, fits the regression curve between each key process parameter and the forming density, and represents its correlation by the slope.

[0080] After determining the key process parameters affecting the density of granular materials during the pressing process, scatter analysis was further carried out to calculate the correlation between key process parameters such as forming pressure, forming temperature, and holding pressure time and the forming density, and a correlation heat map was drawn to visually analyze the correlation between them (as Figure 3 shown).

[0081] After obtaining the types of key parameters in the forming process, a pressing forming test was carried out. Through an orthogonal test with three factors and six levels, the corresponding key process parameters and forming quality data were collected to train the prediction model for the pressing process parameters and forming quality of explosive granular materials. The orthogonal test divides the pressing forming pressure, forming temperature, and holding pressure time, and conducts density tests respectively. The obtained test parameters are used to train the prediction model for the forming quality of explosive granular materials during pressing, fundamentally ensuring the accuracy of the constructed model. The test results collected are shown in Table 1.

[0082] Table 1 Key parameters and density test results in the process of granular material pressing forming (pressing rate 1mm / s)

[0083]

[0084]

[0085] Although process parameters such as pressing temperature, pressing pressure, and holding pressure time are set as constants, there may be abnormal data in the collection of the density of the formed charge column caused by the operator's level and experience. Therefore, the 3σ principle is used to analyze and eliminate abnormal parameters for the density value after the granular material is pressed and formed. Analyze and eliminate abnormal data in the density value that may be caused by the operator's level and experience.

[0086] The abnormal parameter analysis method uses the 3σ principle based on the normal distribution to analyze abnormal parameters for abnormal data that may be caused by the operator's level and experience in the collection of the density of the formed charge column, obtains the qualified data interval after the transformation of the standard normal distribution, and eliminates the abnormal data.

[0087] Since the variance value of the disturbance error is not known in advance, it is necessary to transform the charge density data into the standard normal distribution Z~N(0,1). The specific calculation formula is as follows:

[0088]

[0089] where ρ is the actual density value, is the average value of the charge density, and σ1 is the standard deviation of the charge density data.

[0090] The probability within the interval Z∈(-3,3) is 0.9973. If there is data with a density greater than 3 or less than -3, it is considered that the data is abnormal data.

[0091] The maximum-minimum normalization method is adopted to perform current normalization on the process parameter data without changing the distribution characteristics of the original process parameters. Without changing the distribution characteristics of the original process parameters, the maximum-minimum normalization method avoids the influence of parameters with large dimensions in the model vector on the training and prediction effects, and performs dimension unification, that is, linear normalization of the process parameter data.

[0092] After analyzing and removing the abnormal data of the density parameters during the granule compaction process, it is necessary to normalize the pressing temperature, pressing pressure, and holding time to avoid the algorithm effect caused by huge differences in the dimensions of key parameters. The results of the parameter normalization process are shown in Table 2.

[0093] Table 2 Normalization results of granule compaction process conditions (pressing rate 1mm / s)

[0094]

[0095] The support vector regression model (SVR) is used for training to find a function that is as smooth as possible and minimizes the interval between most data points and the fitting function. The support vector regression model finds a function that is as smooth as possible, minimizes the interval between most data points and the fitting function, and does not exceed the tolerance (the maximum error of the predetermined error bound).

[0096] The support vectors and hyperplane are characterized through the solution process of the Lagrangian function, and the fitting error and the distance from the support vectors to the hyperplane are minimized to construct a smooth function to predict the value of the continuous density. Finally, the optimal solution that minimizes the error and the balance point of the tolerance for the fitting error are found, thereby constructing a model with strong generalization ability. The obtained regression function is:

[0097]

[0098] where a i 、 is the Lagrange multiplier, k(x i , x) represents the kernel function of the support vector regression machine, y j represents the density value in the model, and ε is the tolerance. The Gaussian kernel function is selected as the kernel function of the model.

[0099] The regression model is trained to accurately fit the relationship between the process parameters and the forming density in the data samples. The preprocessed data samples are split into a training set and a test set, and are brought into the support vector machine regression model for training. It is necessary to construct a loss function to measure the fitting effect of the regression model during the training process.

[0100] The Bayesian optimization method defines an optimization objective and uses a Gaussian process model to evaluate the performance function of the hyperfunction, and gradually converges to the optimal hyperparameter value through continuous experiments.

[0101] The Gaussian process refers to a statistical model in which observations are distributed in a continuous space or time. The optimization algorithm uses a Gaussian process probability model.

[0102] The training model can be visualized by plotting the model's predicted value for each point. After dividing the x, y, and z coordinates into 100 equal parts, the model predicts these 103 points respectively. The resulting scatter plot shows the size of the predicted value in color, as shown in the figure below. Figure 4 As shown. Figure 4 It can be found that the higher the molding temperature and molding pressure, the darker the color, and the longer the holding time, the darker the color, but the trend is not so significant. Since the interior of the 100×100×100 cubic graph is not visible, the plane with a holding time of 35 minutes is selected, and the density values ​​of each point on the plane are expressed in a new three-dimensional graph, as shown in Figure 5 In the figure, the x-coordinate is the pressing pressure, the y-coordinate is the pressing temperature, and the z-coordinate is the predicted density value.

[0103] Since the values ​​of model parameters C and γ need to be set in advance, it is difficult to determine their optimal combination. There are many optimization options for RBF kernel parameters (C, γ). The difference in parameters (C, γ) determines the difference in SVR prediction effects. The model parameters need to be optimized. The Bayesian optimization method is used to determine the optimal parameters of the model. The Bayesian formula is:

[0104]

[0105] Among them, A and B represent two random events, P(A i |B) indicates that after the known event B occurs, A i The conditional probability of occurrence is called P(A i |B) is A i The posterior probability, P(B|A i ) is the likelihood function of B; P(A i ) and P(A j ) does not consider the impact of event B, so it is called the prior probability of A.

[0106] Assume that the predicted objective function g(x) obeys the Gaussian process. After the Gaussian process, the new sampling point x is determined by inputting the acquisition function. *, and generate a new Gaussian distribution to obtain the joint Gaussian distribution. According to the calculation formula of the conditional distribution of the multivariate normal distribution, the mean and variance of this conditional distribution can be calculated. Select the acquisition function based on the mean and variance of this conditional distribution, find the sampling point corresponding to the maximum value of the acquisition function, and obtain the observed value of the sampling point, resulting in C = 1.7853 and γ = 0.002451 as the optimal parameters for this optimization process. The parameter optimization flowchart is as Figure 4 shown.

[0107] Test the density of the pressed charge to provide data support for evaluating the conformity between the density calculation result of the quality prediction model and the actual test density result.

[0108] Use a densitometer to test the density of the pressed charge. The test principle of the densitometer is based on "GJB772A-97 Method 401.2 Charge (Block) Density - Hydrostatic Weighing Method". According to the volume of the special immersion liquid displaced by the sample with known mass, the density of the sample is obtained.

[0109] Calculate the average density of the pressed charge after five parallel determinations according to the following formula.

[0110]

[0111] In the above formula, is the average density of the charge, g / cm 3 ; ρ1, ρ2, ρ3, ρ4, and ρ5 are the densities of the pressed charge after five parallel determinations, g / cm 3 .

[0112] Among them, each group of samples is determined five times in parallel and recorded in Table 3.

[0113] Table 3 Test Results of the Density of the Pressed Charge (Pressing Rate is 1 mm / s)

[0114]

[0115] Combined with the analysis of the key parameters in the process of granular compaction, the density test results of the pressed charge, and the density test results of the pressed charge, it is found that the increase in the compaction pressure, forming temperature, and holding time of the granular material can all improve the density of the granular formed charge.

[0116] Use the established prediction model to calculate the density of the formed charge, and analyze the conformity between the calculation result of the formed charge density and the actual density value of the charge tested in the experiment under the same process parameters such as pressing pressure, pressing temperature, holding time, pressing speed, and pressing temperature.

[0117] Prepare the data or dataset for testing, which should be similar to the dataset used in model training, that is, maintain the same number and type of parameters as the training dataset. If it is necessary to have three parameters: pressing pressure, pressing temperature, and holding pressure time, and the order of these three parameters should be the same as the parameter order in the training dataset, as shown in Table 4.

[0118] Table 4 Dataset Saving Format

[0119] A B C 1 Pressure (MPa) Temperature (°C) Pressure holding time (min) 2 83 30 10 3 108 54 30 4 108 65 50 5 125 30 30 6 125 60 20 7 141 30 40

[0120] Load the charge mass prediction model and call the saved support vector regression machine model file.

[0121] Save the prepared dataset to the specified path. After reading the data, convert the data into array data for the model to use.

[0122] Save the density result calculated by the model in the folder with the same path as the input dataset. The calculation result can be viewed by opening this file, where the order of the calculation results corresponds to the order of the input parameters respectively.

[0123] Take the average percentage error (APE) as the evaluation index for the conformity between the calculation result of the prediction model and the test value, which can directly reflect the degree to which the calculation result of the prediction model deviates from the test value and measure the difference between them.

[0124] The calculation method of the average percentage error is shown in the following formula:

[0125]

[0126] Among them, Y i represents the test value of the density of the pressed charge, g / cm 3 ; represents the calculation result of the density of the prediction model, g / cm 3 .

[0127] Use the above formula to calculate the average percentage error between the actual test density value of the formed charge and the density calculation of the pressing process parameters and the forming quality prediction model, and the results are recorded in Table 5.

[0128] Table 5 Average Percentage Error between the Calculation Result of the Quality Prediction Model and the Test Result (Pressing Rate is 1mm / s)

[0129]

[0130] From the evaluation results of the conformity between the calculation results of the above table's quality prediction model and the test results, it is found that under different pressing and forming process conditions, the average percentage error between the calculated value of the charge density by the pressing process parameters and the forming quality prediction model and the average density value of the actually measured charge is at least 3.25% (conformity 96.75%) and at most 8.03% (conformity 91.97%). The results show that the accuracy of the prediction model of the pressing process parameters and the forming quality established by this method is relatively high.

[0131] In summary, the method for predicting the quality of particulate matter pressing and forming provided by this application constructs a prediction model based on the correlation between key process parameters and the density of the formed charge. It can predict and analyze the density and density distribution of particulate matter after pressing and forming, effectively making up for the deficiency of the current actual characterization means in characterizing the characteristic parameters of the forming process of particulate matter in a dense and non-transparent restricted space. It can provide intuitive and accurate real-time data support for the study of the variation laws of the temperature field and stress field of particulate matter during the pressing and forming process under different process conditions, and lay a theoretical foundation for the precise control of the forming process parameters and the further improvement of the forming quality.

[0132] See Figure 6 , the embodiment of this application can also provide a device for predicting the quality of particulate matter pressing and forming. As Figure 6 shown, it is used to execute the above-mentioned method for predicting the quality of particulate matter pressing and forming. The device may include:

[0133] A key parameter determination unit 601, which is used to transform the variables with high correlation during the particulate matter pressing and forming process into variables that are independent of each other or have low correlation by using the principal component analysis method, so as to determine three key parameters required for constructing the forming quality prediction model;

[0134] A correlation calculation unit 602, which is used to determine the correlation between each of the key parameters and the forming density by using the scatter plot analysis method;

[0135] A density data acquisition unit 603, which is used to obtain the density data of the formed charge through an orthogonal experiment with three factors and six levels by combining the three key parameters;

[0136] An abnormal data rejection unit 604, which is used to reject the abnormal data generated during the data acquisition process according to the 3σ principle by using the abnormal parameter analysis method;

[0137] A sample data acquisition unit 605, which is used to unify the data dimension without changing the original distribution characteristics of the process parameters by using the maximum-minimum normalization method to obtain sample data;

[0138] A model construction unit 606 is configured to characterize support vectors and hyperplanes through the solution process of the Lagrangian function, and minimize the fitting error and the distance from the support vectors to the hyperplane, so as to construct and obtain a support vector regression machine model, where the support vector regression machine model includes a smoothing function for predicting continuous density values;

[0139] A model optimization unit 607 is configured to construct a loss function in a regression problem to measure the fitting effect of the support vector regression machine model during training, input the sample data for training the support vector regression machine model, and optimize the model performance by using a Bayesian optimization method to obtain optimal parameters, so as to obtain the molding quality prediction model;

[0140] A quality prediction unit 608 is configured to input test data into the molding quality prediction model, so that the molding quality prediction model outputs a molding quality prediction result; the test data includes at least three test parameters, and the types and orders of the three test parameters are the same as those of the three key parameters in the sample data.

[0141] An embodiment of the present application may further provide a granular material molding quality prediction device, where the device includes a processor and a memory:

[0142] The memory is configured to store program code and transmit the program code to the processor;

[0143] The processor is configured to execute the steps of the above-mentioned granular material molding quality prediction method according to the instructions in the program code.

[0144] As Figure 7 shown, a granular material molding quality prediction device provided by an embodiment of the present application may include: a processor 10, a memory 11, a communication interface 12, and a communication bus 13. The processor 10, the memory 11, and the communication interface 12 all complete communication with each other through the communication bus 13.

[0145] In an embodiment of the present application, the processor 10 may be a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic devices, etc.

[0146] The processor 10 may call the program stored in the memory 11. Specifically, the processor 10 may execute the operations in the embodiment of the granular material molding quality prediction method.

[0147] The memory 11 is used to store one or more programs, which may include program codes, and the program codes include computer operation instructions. In the embodiment of the present application, the memory 11 at least stores programs for implementing the following functions:

[0148] The principal component analysis method is used to transform the variables with high correlation in the granular pressing process into independent or low-correlation variables to determine the three key parameters required to build a molding quality prediction model.

[0149] The correlation between each of the key parameters and the molding density is determined by using a scatter analysis method;

[0150] The density data of the formed drug column is obtained by combining the three key parameters through a three-factor six-level orthogonal test;

[0151] The abnormal parameter analysis method is used to eliminate the abnormal data generated during the data collection process according to the 3σ principle;

[0152] The maximum and minimum normalization method is used to unify the data dimension and obtain sample data without changing the distribution characteristics of the original process parameters;

[0153] Characterizing the support vector and the hyperplane through the Lagrangian function solution process, and minimizing the fitting error and the distance from the support vector to the hyperplane, so as to construct a support vector regression machine model, wherein the support vector regression machine model includes a smoothing function for predicting continuous density values;

[0154] Constructing a loss function in the regression problem to measure the fitting effect of the support vector regression machine model during the training process, bringing the sample data into the training of the support vector regression machine model, optimizing the model performance by using the Bayesian optimization method, and obtaining the optimal parameters, so as to obtain the molding quality prediction model;

[0155] Input test data into the molding quality prediction model so that the molding quality prediction model outputs a molding quality prediction result; the test data includes at least three test parameters, and the types and order of the three test parameters are the same as those of the three key parameters in the sample data.

[0156] In one possible implementation, the memory 11 may include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required for at least one function (such as a file creation function, a data reading and writing function), etc.; the data storage area can store data created during use, such as initialization data, etc.

[0157] In addition, the memory 11 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0158] The communication interface 12 can be an interface of the communication module for connecting to other devices or systems.

[0159] Of course, it should be noted that Figure 7 the structure shown does not constitute a limitation on the granular material compacting quality prediction device in the embodiments of the present application. In actual applications, the granular material compacting quality prediction device may include more or fewer components than those shown, or combine certain components. Figure 7 than those shown, or combine certain components.

[0160] The embodiments of the present application can also provide a computer-readable storage medium for storing program codes for executing the steps of the above-mentioned granular material compacting quality prediction method.

[0161] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0162] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this 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, which can be stored in a storage medium such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0163] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for a system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for the relevant parts, reference can be made to the description of the method embodiment. The systems and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0164] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for predicting the quality of granular body compaction, characterized in that: include: The principal component analysis method is used to transform the variables with high correlation in the granular pressing process into independent or low-correlation variables to determine the three key parameters required to build a molding quality prediction model. The correlation between each of the key parameters and the molding density is determined by using a scatter analysis method; The density data of the formed drug column is obtained by combining the three key parameters through a three-factor six-level orthogonal test; The abnormal parameter analysis method is used to eliminate the abnormal data generated during the data collection process according to the 3σ principle; The maximum and minimum normalization method is used to unify the data dimension and obtain sample data without changing the distribution characteristics of the original process parameters; Characterizing the support vector and the hyperplane through the Lagrangian function solution process, and minimizing the fitting error and the distance from the support vector to the hyperplane, so as to construct a support vector regression machine model, wherein the support vector regression machine model includes a smoothing function for predicting continuous density values; Constructing a loss function in the regression problem to measure the fitting effect of the support vector regression machine model during the training process, bringing the sample data into the training of the support vector regression machine model, optimizing the model performance by using the Bayesian optimization method, and obtaining the optimal parameters, so as to obtain the molding quality prediction model; Test data is input into the molding quality prediction model so that the molding quality prediction model outputs a molding quality prediction result; the test data includes at least three test parameters, and the types and order of the three test parameters are the same as those of the three key parameters in the sample data.

2. The method for predicting the quality of granular body pressing according to claim 1, characterized in that: The principal component analysis method includes calculating the principal component cumulative contribution rate formula to obtain the principal component contribution rate of each process parameter to the molding density, and the contribution rate is calculated by the following formula: Where: Q (p) represents the cumulative contribution rate of the first p principal components, λ k Represents the eigenvalue.

3. The method for predicting the quality of granular body pressing according to claim 2, characterized in that: The three key parameters include molding pressure, molding temperature and holding time.

4. The method for predicting the quality of granular body compacting according to claim 1, characterized in that: The scatter analysis method used to determine the correlation between each of the key parameters and the molding density includes: Each variable is normalized to eliminate the dimensional differences between the variables, and a regression curve between each key parameter and the molding density is fitted, and the correlation with the molding density is expressed by the slope.

5. The method for predicting the quality of granular body compacting according to claim 1, characterized in that: The kernel function of the support vector regression model is a Gaussian kernel function.

6. The method for predicting the quality of granular body compacting according to claim 5, characterized in that: The support vector regression model is represented by the following formula: Where: a i , is the Lagrange multiplier, k(x i , x) represents the kernel function of the support vector regression machine, y j Represents the density value in the model, and ε is the tolerance.

7. The method for predicting the quality of granular body compacting according to claim 1, characterized in that: The Bayesian optimization method is expressed as follows: Where: A and B represent two random events, P(A i |B) indicates that after the known event B occurs, A i The conditional probability of occurrence, that is, P(A i |B) is A i The posterior probability, P(B|A i ) is the likelihood function of B; P(A i ) and P(A j ) does not consider the impact of event B on the prior probability of A.

8. A device for predicting the quality of granular body compaction, characterized in that: The device is used to execute the method for predicting the quality of granular body pressing molding according to any one of claims 1 to 7, and comprises: A key parameter determination unit is used to transform highly correlated variables in the granular body pressing process into independent or low-correlated variables using a principal component analysis method, so as to determine three key parameters required for building a molding quality prediction model; A correlation calculation unit, used for determining the correlation between each of the key parameters and the molding density by using a scatter point analysis method; A density data acquisition unit, used to obtain density data of the formed drug column by combining the three key parameters through a three-factor six-level orthogonal test; An abnormal data elimination unit is used to eliminate abnormal data generated during the data collection process using an abnormal parameter analysis method according to the 3σ principle; A sample data acquisition unit is used to obtain sample data by unifying the data dimension without changing the original process parameter distribution characteristics by using the maximum and minimum normalization method; A model building unit, used for characterizing support vectors and hyperplanes through a Lagrangian function solution process, and minimizing fitting errors and distances from support vectors to hyperplanes, so as to construct a support vector regression machine model, wherein the support vector regression machine model includes a smoothing function for predicting continuous density values; A model optimization unit, used for constructing a loss function in the regression problem to measure the fitting effect of the support vector regression machine model during the training process, bringing in the sample data to train the support vector regression machine model, optimizing the model performance by using the Bayesian optimization method, and obtaining the optimal parameters, so as to obtain the molding quality prediction model; The quality prediction unit is used to input test data into the molding quality prediction model so that the molding quality prediction model outputs a molding quality prediction result; the test data includes at least three test parameters, and the types and order of the three test parameters are the same as those of the three key parameters in the sample data.

9. A device for predicting the quality of granular body pressing, characterized in that: The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the bulk body pressing quality prediction method described in any one of claims 1 to 7 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and the program code is used to execute the method for predicting the quality of granular body pressing molding according to any one of claims 1-7.

Images