CN113255083B - Elastic modulus prediction method based on molecular dynamics and elastic network regression model - Google Patents
Elastic modulus prediction method based on molecular dynamics and elastic network regression model Download PDFInfo
- Publication number
- CN113255083B CN113255083B CN202110715862.2A CN202110715862A CN113255083B CN 113255083 B CN113255083 B CN 113255083B CN 202110715862 A CN202110715862 A CN 202110715862A CN 113255083 B CN113255083 B CN 113255083B
- Authority
- CN
- China
- Prior art keywords
- elastic modulus
- elastic
- constructing
- glass
- cation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Abstract
The invention discloses an elastic modulus prediction method based on molecular dynamics and an elastic network regression model, and belongs to the field of glass performance prediction. According to the method, crystal and amorphous structures of oxides in the glass are established based on molecular dynamics, long-range disorder and short-range order of the glass are considered, the method is more consistent with the rule of actual glass, and aiming at the property of the glass, namely the elastic modulus, the binding energy of unit cations in the crystal, the elasticity tensor constant of a unit cell and the product of the body modulus and the shear modulus of the unit cell are innovatively provided as performance parameters, and descriptors used in an elastic network regression model are correspondingly constructed, so that the accurate prediction of the elastic modulus of the glass is realized; and based on the molecular dynamics calculation descriptor and Newton's law, the system is simple in structure, the action of each electron does not need to be considered, but the electronic mechanism of the system is equivalent to the potential energy function in the atom, so that the calculation process is greatly simplified, and the calculation time is saved.
Description
Technical Field
The invention relates to an elastic modulus prediction method based on molecular dynamics and an elastic network regression model, and belongs to the field of glass performance prediction.
Background
The elastic modulus is an important performance parameter of engineering materials, and is a measure for the resistance of an object to elastic deformation in a macroscopic view and is a reflection of the bonding strength between atoms, ions or molecules in a microscopic view. All factors influencing the bonding strength can influence the elastic modulus of the material, such as bonding mode, crystal structure, chemical composition, microstructure, temperature and the like. The elastic modulus can be regarded as an index for measuring the difficulty of the material in elastic deformation, and the larger the value of the elastic modulus, the larger the stress for causing the material to generate certain elastic deformation, that is, the higher the rigidity of the material, that is, the smaller the elastic deformation generated under the action of certain stress.
The modulus of elasticity of the material is very important in engineering design, for example, in the use of wind power blades, the blade mass is also increased significantly as the length of the blade is increased. The substantial increase in mass can present a number of challenges to the design of the blade: for example, the risk of layering between the main beam cap and the skin is greatly increased, the natural frequency and the rotating frequency of the blade are closer, the transportation and hoisting difficulty of the blade is greatly increased, and the unit needs to bear larger load. Therefore, the current research focus is on improving the structural rigidity of the wind turbine blade spar cap, particularly reflecting on the raw materials, and improving the tensile modulus of the used glass fiber.
Due to the appearance of big data, a plurality of methods for predicting material performance based on a machine learning algorithm are available at present, the methods establish a prediction model based on a neural network or a support vector machine and train and verify the prediction model to obtain a model under optimal parameters, but the methods predict the percentage of material components as input, so that the prediction accuracy is not high, a method for predicting the glass performance by using a descriptor calculated by molecular dynamics as the input of the machine learning algorithm is not available at present, a method for accurately predicting the elastic modulus by using the machine learning algorithm of the molecular dynamics is not provided aiming at an influence factor of the elastic modulus, and the method for predicting the elastic modulus is a key problem to be solved urgently at present.
Disclosure of Invention
In order to solve the problem that the elastic modulus of glass cannot be intuitively, quickly and accurately predicted in the prior art, the invention provides an elastic modulus prediction method based on molecular dynamics and an elastic network regression model, which comprises the following steps:
step 1, acquiring elastic modulus data of oxide glass materials with different component compositions, and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
step 2, constructing an atomic structure model of oxide crystals with different symmetries in the oxide glass material based on molecular dynamics, and using cations per unitiBinding energy ofElastic tensor constant of each unit cell of cubic crystal containing cation AAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusConstructing a descriptor of the elastic modulus comprising a "material gene" as a performance parameter;
step 3, constructing a training set, a verification set and a test set based on the elastic modulus database constructed in the step 1 and the descriptor constructed in the step 2;
step 4, constructing an elastic modulus prediction model based on the elastic network regression model, and training the constructed elastic modulus prediction model according to the training set, the verification set and the test set constructed in the step 3 to obtain a trained elastic modulus prediction model;
step 5, aiming at the glass material to be predicted, obtaining a descriptor containing a material gene according to the composition of the components of the glass material; and predicting the elastic modulus of the glass material to be predicted by using the trained elastic modulus prediction model.
Optionally, step 2 includes:
step 2-1, constructing atomic structure models of oxide crystals with different symmetries as unit cells of molecular dynamics calculation;
step 2-2 for each cell structure constructed in step 2-1, performing molecular dynamics calculation to obtain cell energy;
Step 2-3, obtaining a performance parameter set of each crystal cell structure obtained in the step 2-1 through further calculation, and constructing a descriptor for machine learning:
In the formula (I), the compound is shown in the specification,nis a power-averaged exponent, and takes on all non-zero integers between-4 and + 4;is the set of all cations;is a corresponding cationiThe component ratio of (A);is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,3;is a power exponent ofWhen is a cationIs corresponding tomA descriptor; the performance parameters include the cation per unit of each crystaliBinding energy ofElastic tensor constant of each unit cell of cubic crystal containing cation AAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a).
Optionally, the cation per unitiBinding energy ofThe calculation method comprises the following steps:
using the cell energy obtained in step 2-2Subtracting the sum of the energies of the single atoms of the same number and kind, and calculating by the formula:
in the formula (I), the compound is shown in the specification,Mfor cations in crystal cellsiThe number of the (c) component (a),andare each a single cationiAnd a single oxygen atom, O, in one unit cell.
Alternatively, the step 2-1 of constructing an atomic structure model of oxide crystals having different symmetries as a unit cell of molecular dynamics calculation includes:
(1) for cations with a valence of +1, according to CaF2The structure is crystal, oxygen ions form a face-centered cubic structure, and cations are in tetrahedral gaps of the face-centered cubic structure;
(2) for + 2-valent cations, crystals are constructed according to a NaCl structure, the cations form a face-centered cubic structure, and oxygen ions are in hexahedral gaps of the face-centered cubic structure;
(3) for cations with a valence of +3, according to α -Al2O3A structural structure crystal;
(4) for cations with a valence of +4, according to CaF2The structure makes crystal, the positive ions form a face-centered cubic structure, and the oxygen ions are in the tetrahedral gaps of the face-centered cubic structure.
Optionally, when the elastic modulus database is constructed in step 1, the method further includes preprocessing the elastic modulus data of the oxide glass materials with different components:
judging whether the following two conditions are simultaneously satisfied or not according to the two glass components:
condition 1: in the glass components, the molar ratio difference value of the components of each oxide component is less than or equal to a first preset threshold, and the unit is percentage;
condition 2: the difference value of the elastic moduli of the two glasses is greater than a second preset threshold value, and the unit is percentage;
and if the two glass components are simultaneously satisfied, removing the corresponding two glass components and the corresponding elastic modulus data from the elastic modulus database.
Optionally, the first preset threshold is 2%, and the second preset threshold is 20%.
Optionally, in the elastic network regression model, the elastic modulus is predictedyAnd descriptorsThe following relationship is satisfied:
wherein the content of the first and second substances,Perin order to be a set of performance parameters,is composed ofCorresponding regression coefficients, n being a non-zero integer from-4 to 4;
the loss function of the elastic network regression model is:
wherein the content of the first and second substances,for the loss function, N is the number of samples,in order to predict the value of the target,in the form of an actual value of the value,in order to be a penalty term,for adjusting the over-parameters;
The present application further provides an elastic modulus prediction system based on molecular dynamics and elastic network regression algorithm, the system comprising:
the elastic modulus database construction module is used for acquiring elastic modulus data of glass materials with different component compositions and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
a descriptor construction module for constructing a descriptor containing a 'material gene' of the elastic modulus of the oxide glass material based on molecular dynamics;
the training data construction module is used for constructing a training set, a verification set and a test set on the basis of the elastic modulus database constructed by the elastic modulus database construction module and the descriptor constructed by the descriptor construction module;
the model construction training module is used for constructing an elastic modulus prediction model based on an elastic network regression model tree, and training the constructed elastic modulus prediction model according to a training set, a verification set and a test set constructed by the training data construction module to obtain a trained elastic modulus prediction model;
and the prediction module is used for constructing an elastic modulus prediction model trained by the training module by utilizing the model to predict the elastic modulus of the glass material to be predicted.
Optionally, the descriptor building module constructs a descriptor containing a "material gene" of the elastic modulus of the oxide glass material, including
Constructing an atomic structure model of oxide crystals with different symmetries as a unit cell for molecular dynamics calculation;
for each cell structure obtained by construction, molecular dynamics calculation is carried out to obtain unit cell energy;
For each cell structure obtained by construction, a performance parameter set is obtained by further calculationConstructing descriptors for machine learning:
In the formula (I), the compound is shown in the specification,nis a power-averaged exponent, and takes on all non-zero integers between-4 and + 4;is the set of all cations;is a corresponding cationiThe component ratio of (A);is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,3;is a power exponent ofWhen is a cationThe m-th descriptor corresponding to the structure of (1); the performance parameters include the cation per unit of each crystaliBinding energy ofElastic tensor constant of each unit cell of cubic crystal containing cation AAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a).
The present application further provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.
The invention has the beneficial effects that:
1) the crystal and amorphous structures of the oxides in the glass are established, the long-range disorder and short-range order of the glass are considered, the rule of the actual glass is better met, and the calculation result is accurate.
2) Based on the molecular dynamics calculation descriptor and Newton's law, the system is simple in structure, the action of each electron does not need to be considered, the electronic mechanism of the system is equivalent to the potential energy function in the atom, the calculation process is greatly simplified, and the calculation time is saved.
3) By adopting the elastic network regression model, global linear regression does not need to be considered for samples with more sample characteristics and complex nonlinear relations among the characteristics, the elastic network regression model divides the sample characteristics into a plurality of piecewise linear regressions, the interpretability of the tree is stronger, and the training model time of the algorithm is short.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a method for predicting glass modulus based on molecular dynamics in one embodiment.
FIG. 2 is a flow diagram for constructing descriptors based on molecular dynamics, in one embodiment.
FIG. 3 shows CaF2Structural diagram of crystal form.
FIG. 4 is a view showing a structure of NaCl type crystals.
FIG. 5 shows a-Al2O3Structural diagram of crystal form.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The first embodiment is as follows:
the present embodiment provides an elastic modulus prediction method based on molecular dynamics and elastic network regression algorithm, referring to fig. 1, the method includes:
step 1, acquiring elastic modulus data of glass materials with different component compositions, and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
step 2, constructing descriptors containing 'material genes' of the elastic modulus of the oxide glass material based on molecular dynamics, and referring to fig. 2;
2-1 constructing an atomic structure model of oxide crystals with different symmetries as a unit cell for molecular dynamics calculation;
2-2 for each crystal cell structure constructed in the step 2-1, performing molecular dynamics calculation to obtain optimized unit cell energy;
2-3 are directed to the structures of the above crystals comprising various symmetries optimized by molecular dynamics calculations; and further calculating to obtain a performance parameter set thereof, and constructing a descriptor for machine learning:
In the formula (I), the compound is shown in the specification,nis a power-averaged exponent, takes on all non-zero integers between-4 and +4,is the set of all the cations that are present,is a corresponding cationiThe ratio of the components (A) to (B),is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,is a power exponent ofWhen is a cationIs corresponding tomA descriptor.
The set of performance parameters includes the following performance parameters:
calculation of binding energy by using optimized unit cell energySubtracting the sum of the energies of the single atoms with the same number and the same type to obtain the product, wherein the calculation formula is as follows:
in the formula (I), the compound is shown in the specification,Mfor cations in crystal cellsiThe number of the (c) component (a),andare each a single cationiAnd the energy of a single oxygen atom O in one unit cell can be calculated through the command statements of computer ke/atom and computer pe/atom of Lammps software;
(3) volume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a);
step 3, constructing a training set, a verification set and a test set based on the elastic modulus database constructed in the step 1 and the descriptor constructed in the step 2;
step 4, constructing an elastic modulus prediction model based on the elastic network regression model, and training the constructed elastic modulus prediction model according to the training set, the verification set and the test set constructed in the step 3 to obtain a trained elastic modulus prediction model;
and 5, aiming at the glass material to be predicted, predicting the elastic modulus of the glass material by using the trained elastic modulus prediction model.
Example two:
the embodiment provides an elastic modulus prediction method based on molecular dynamics and elastic network regression algorithm, which includes:
step 1, acquiring elastic modulus data of glass materials with different component compositions, and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
for the acquired elastic modulus data, in practical application, the method further comprises preprocessing the acquired elastic modulus data, wherein the preprocessing comprises the following steps:
judging whether the following two conditions are simultaneously satisfied or not according to the two glass components:
condition 1: in the glass components, the molar ratio difference value of the components of each oxide component is less than or equal to a first preset threshold, and the unit is percentage;
condition 2: the difference value of the elastic moduli of the two glasses is greater than a second preset threshold value, and the unit is percentage;
if the two glass components are simultaneously satisfied, the corresponding two glass components and the corresponding elastic modulus data are removed from the database.
Wherein, the first preset threshold and the second preset threshold in the above two conditions can be determined by the technicians in the field according to the prior knowledge, in the present application, the first preset threshold is 2%, and the second preset threshold is 20%
Step 2, constructing a descriptor containing a material gene of the elastic modulus of the oxide glass material based on molecular dynamics;
2-1 constructing an atomic structure model of oxide crystals with different symmetries as a unit cell for molecular dynamics calculation;
specifically, the method comprises the following steps:
(1) for cations with a valence of +1, according to CaF2The structure is crystal, oxygen ions form a face-centered cubic structure, and cations are in tetrahedral gaps of the face-centered cubic structure, as shown in figure 3;
(2) for +2 valent cations, crystals are constructed according to the NaCl structure, the cations form a face-centered cubic structure, and oxygen ions are in hexahedral gaps of the face-centered cubic structure, as shown in FIG. 4;
(3) for cations with a valence of +3, according to α -Al2O3A structural structure crystal, as shown in fig. 5;
(4) for cations with a valence of +4, according to CaF2The structure makes crystal, the positive ions form a face-centered cubic structure, and the oxygen ions are in the tetrahedral gaps of the face-centered cubic structure;
2-2 for each crystal cell structure constructed in the step 2-1, performing molecular dynamics calculation to obtain optimized unit cell energy;
Specifically, the method can be realized by adopting the software LAMMPS of molecular dynamics calculation, and the optimized unit cell energy is obtained according to the following steps:
(a) Selecting a potential function of the system as a Morse potential:
the parameters used for the different oxide elements are shown in the following table:
table 1: parameters adopted by Morse potential
(eV) | () | () | (eV) | |
Li-O | 0.001114 | 3.429506 | 2.681360 | 1.0 |
Na-O | 0.023363 | 1.763867 | 3.006315 | 5.0 |
K-O | 0.011162 | 2.062605 | 3.305308 | 5.0 |
Mg-O | 0.038908 | 2.281000 | 2.586153 | 5.0 |
Ca-O | 0.030211 | 2.241334 | 2.923245 | 5.0 |
Ba-O | 0.065011 | 1.547596 | 3.393410 | 5.0 |
Si-O | 0.340554 | 2.006700 | 2.100000 | 0.9 |
Al-O | 0.361581 | 1.900442 | 2.164818 | 1.0 |
Ti-O | 0.024235 | 2.254703 | 2.708943 | 1.0 |
Sr-O | 0.019623 | 1.886000 | 3.328330 | 3.0 |
(b) Determining initial conditions, given an initial configuration and an initial velocity of the particles (a reasonable initial configuration may accelerate the system towards equilibrium), and according to the boltzmann distribution, assigning initial velocities (v) to the atoms:
(c) the energy conservation of the system is controlled during the simulation using a micro-canonical ensemble (NEV). The temperature and pressure eventually reach equilibrium.
(d) The time step length adopted by simulation is 1fs, the simulation time is 30ps, and the truncation radius is 10。
(e) After the system reaches the equilibrium, the configuration integral of the system is solved through the tracks (r, F) along the phase space, and then the energy of the unit cell is further calculated based on the result of the configuration integral。
(f) Single cation is calculated through the command statements of computer ke/atom and computer pe/atom of Lammps softwareiAnd a single oxygen atom O in a unit cellAnd。
2-3 are directed to the structures of the above crystals comprising various symmetries optimized by molecular dynamics calculations; and further calculating to obtain a performance parameter set thereof, and constructing a descriptor for machine learning:
In the formula (I), the compound is shown in the specification,nis a power average exponentAll non-zero integers between-4 and +4,is the set of all the cations that are present,is a corresponding cationiThe ratio of the components (A) to (B),is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,is a power exponent ofWhen is a cationIs corresponding tomA descriptor.
Cationic in predicting elastic modulus hereinThe structures of (1) correspond to 3 performance parameters and three performance parameters respectivelyRespectively as follows:
calculation of binding energy by using optimized unit cell energySubtracting the sum of the energies of the single atoms with the same number and the same type to obtain the product, wherein the calculation formula is as follows:
in the formula (I), the compound is shown in the specification,Mfor cations in crystal cellsThe number of the (c) component (a),andare each a single cationAnd the energy of a single oxygen atom, O, in a unit cell;
(3) volume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a);
the ElaStic tensor constant of the hexagonal crystal can be calculated through the tool software ElaStac of Quantum Espresso,,,And. Volume of the hexagonal crystalAnd Voigt average shear modulusThe calculation formula is as follows:
using a set of data in a database as an example, the application is introduced to construct descriptors for machine learningThe process of (2) is as follows:
a group collected from the databaseThe data are as follows: the glass structure contains A mol SiO2,B mol Al2O3,C mol Na2O, the elastic modulus of the component glass is y GPa.
The total number of ions was calculated to be (3A + 5B + 3C) mol, where Si4+Proportion of ionsComprises the following steps:
Al3+the proportion is as follows:
Na+the proportion is as follows:
when predicting the modulus of elasticity, for each cationi(Si,Al,Na),All consist of the 3 types of parameters mentioned above:elastic constant ofAndmodulus parameter (body modulus)And shear modulusThe product of (d).
n =1 is as follows:
n =2 as follows:
by analogy, the rest can be constructednDescriptor (n = -4, -3, -2, -1, 3, 4) case of value (each case)nThe value has 3 descriptors).
Step 3, constructing a training set, a verification set and a test set based on the elastic modulus database constructed in the step 1 and the descriptor constructed in the step 2;
the method specifically comprises the following steps:
step 3-1, from the elastic modulus database according to the total data quantity NRandomly extracting data as a first test subset { T }1};
Step 3-2, for the restScreening out data with elastic modulus value smaller than third preset threshold value, and randomly selecting dataAs a second test subset { T }2};
Step 3-3, test subset for divide { T }1}、{T2Acquiring the residual data, acquiring the glass components of the concerned specific components in a preset section, and then randomly selecting the glass components from the glass componentsAs a third test subset { T }3};
And 3-4, combining the three test subsets to form a test set { T } = { T } of the model1,T2,T3The rest data are used as a training set and a verification set of the model, and the data are used as the training set and the verification set of the model、Andthe value of (A) is to ensure that the data ratio of the training set plus the verification set ({ D } + { V }) to the test set { T } is 9: 1.
And 3-4, taking the rest data as a training set and a verification set of the model, wherein the specific division process comprises the following steps:
constructing a training set { D } and a verification set { V } by adopting a k-fold cross validation method, and specifically comprising the following steps:
step 3-4-1, arranging the rest 90% N data in the database according to the elastic modulus value in an ascending order, and then averagely dividing the data into k disjoint subsets { S1,S2,S3……Sk};
Step 3-4-2, 1 subset S of the data is taken each timehAs a verification set { V }hH =1, 2, 3 … … k, the remaining k-1 subsets as training set { D }hWill train set { DhV and verification sethAs cross validation data.
Step 4, constructing an elastic modulus prediction model based on the elastic network regression model, and training the constructed elastic modulus prediction model according to the training set, the verification set and the test set constructed in the step 3 to obtain a trained elastic modulus prediction model;
step 4-1, establishing an elastic network regression model, and setting a predicted elastic modulus y and a descriptorThe following relationship is satisfied:
whereinPerIn order to be a set of performance parameters,is composed ofCorresponding regression coefficients, n being a non-zero integer from-4 to 4;
step 4-2, establishing a loss function of the elastic network regression model according to the following formula:
wherein the content of the first and second substances,for the loss function, N is the number of samples,in order to predict the value of the target,in the form of an actual value of the value,in order to be a penalty term,for adjusting the over-parameters;
Step 4-3, adjusting the hyper-parameters for a groupCoefficient of performanceSolving for minimizationOptimization of lower correspondenceHyper-parameters;
the specific process comprises the following steps:
step 4-3-1, inUnder the condition, the training set { D obtained in the step 3 is adoptedhV and verification sethTraining models in turnkThen, and calculatekNext-in verification set { VhMean square error MSE value in (j):
in the formula (I), the compound is shown in the specification,in order to verify the amount of data in a set,in order to verify the predictive value of a set,actual values for the validation set;
Step 4-3-3, adjusting model parameters of elastic network regressionSetting the adjusting parameters in sequenceCoefficient ofTo adjust the over-parametersCoefficient ofRepeating the steps 4-3-1 to 4-3-2, and calculating the corresponding;
Step 4-4, all training sets { D in the database are takenhV and verification sethThe sum of the parameters is used as a training set of the final model after parameter adjustment is finished, namely { S }1,S2,S3……Sk};
Step 4-5, training set { S ] based on the elastic network regression model in step 4-11,S2,S3……SkAfter training, a series of regression coefficients of the model are obtainedAnd forming an elastic modulus prediction model.
And 5, aiming at the glass material to be predicted, predicting the elastic modulus of the glass material by using a trained elastic modulus prediction model, wherein the specific process comprises the following steps of:
step 5-1, according to the process of step 2, constructing a descriptor of the glass material to be predicted;
Step 5-2, the descriptor is processedAnd substituting the model into the elastic modulus prediction model to obtain the predicted elastic modulus.
To verify the effectiveness of the method of the present application, 10 sets of glass materials with known values of elastic modulus were predicted by the method of the present invention, and the results are shown in table 2 below.
Table 2: comparison of predicted value and true value of glass elastic modulus predicted by the method
As can be seen from the above table 2, the maximum error between the elastic modulus value of the glass predicted by the method provided by the invention and the true value is only 5.4%, and compared with the existing method, the method provided by the invention can relatively accurately predict the elastic modulus value and verify the effectiveness of the method. Moreover, the method of the invention is adopted to predict the elastic modulus of the glass with unknown elastic modulus, can quickly estimate the elastic modulus of the glass with different component proportions, can greatly reduce the trial and error cost for glass research and development, and has great significance for research and development targets with strict requirements on the elastic modulus of the glass.
Example three:
the embodiment provides an elastic modulus prediction system based on molecular dynamics and elastic network regression algorithm, the system includes:
the elastic modulus database construction module is used for acquiring elastic modulus data of glass materials with different component compositions and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
a descriptor construction module for constructing a descriptor containing a 'material gene' of the elastic modulus of the oxide glass material based on molecular dynamics;
the training data construction module is used for constructing a training set, a verification set and a test set on the basis of the elastic modulus database constructed by the elastic modulus database construction module and the descriptor constructed by the descriptor construction module;
the model construction training module is used for constructing an elastic modulus prediction model based on an elastic network regression model, and training the constructed elastic modulus prediction model according to a training set, a verification set and a test set constructed by the training data construction module to obtain a trained elastic modulus prediction model;
and the prediction module is used for constructing an elastic modulus prediction model trained by the training module by utilizing the model to predict the elastic modulus of the glass material to be predicted.
For specific definition of the elastic modulus prediction system based on molecular dynamics, reference may be made to the definition of the elastic modulus prediction method based on molecular dynamics, which is not described herein again.
The modules in the above-described system for predicting elastic modulus based on molecular dynamics can be implemented in whole or in part by software, hardware, and combinations thereof.
The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
Example four:
the embodiment provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the following steps:
step 1, acquiring elastic modulus data of a glass material, and constructing an elastic modulus database, wherein the database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
step 2, constructing a descriptor containing a material gene of the elastic modulus of the oxide glass material based on molecular dynamics;
step 3, constructing a training set, a verification set and a test set based on the elastic modulus database and the descriptor constructed in the step 2;
step 4, constructing an elastic modulus prediction model based on an elastic network regression model;
and 5, aiming at the glass material to be predicted, predicting the elastic modulus of the glass material by using the elastic modulus prediction model.
For the specific definition of each step, reference may be made to the definition of the elastic modulus prediction method based on molecular dynamics, which is not described herein again.
In one embodiment, a computer-readable storage medium is provided, having stored thereon a computer program which, when executed by a processor, performs the steps of:
step 1, acquiring elastic modulus data of a glass material, and constructing an elastic modulus database, wherein the database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
step 2, constructing a descriptor containing a material gene of the elastic modulus of the oxide glass material based on molecular dynamics;
step 3, constructing a training set, a verification set and a test set based on the elastic modulus database and the descriptor constructed in the step 2;
step 4, constructing an elastic modulus prediction model based on the elastic network regression model;
and 5, aiming at the glass material to be predicted, predicting the elastic modulus of the glass material by using the elastic modulus prediction model.
For the specific definition of each step, reference may be made to the definition of the elastic modulus prediction method based on molecular dynamics, which is not described herein again.
Some steps in the embodiments of the present invention may be implemented by software, and the corresponding software program may be stored in a readable storage medium, such as an optical disc or a hard disk.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A method for predicting elastic modulus based on molecular dynamics and elastic network regression models, the method comprising:
step 1, acquiring elastic modulus data of oxide glass materials with different component compositions, and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
step 2, constructing atomic structures of oxide crystals with different symmetries in the oxide glass material based on molecular dynamicsModel and in terms of cations per unitiBinding energy ofElastic tensor constant of each unit cell of cubic crystal containing cation AAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusConstructing a descriptor of the elastic modulus comprising a "material gene" as a performance parameter;
step 3, constructing a training set, a verification set and a test set based on the elastic modulus database constructed in the step 1 and the descriptor constructed in the step 2;
step 4, constructing an elastic modulus prediction model based on the elastic network regression model, and training the constructed elastic modulus prediction model according to the training set, the verification set and the test set constructed in the step 3 to obtain a trained elastic modulus prediction model;
step 5, aiming at the glass material to be predicted, obtaining a descriptor containing a material gene according to the composition of the components of the glass material; and predicting the elastic modulus of the glass material to be predicted by using the trained elastic modulus prediction model.
2. The method of claim 1, wherein the step 2 comprises:
step 2-1, constructing atomic structure models of oxide crystals with different symmetries as unit cells of molecular dynamics calculation;
step 2-2 for each cell structure constructed in step 2-1, performing molecular dynamics calculation to obtain cell energy;
Step 2-3, obtaining a performance parameter set of each crystal cell structure obtained in the step 2-1 through further calculation, and constructing a descriptor for machine learning:
In the formula (I), the compound is shown in the specification,nis a power-averaged exponent, and takes on all non-zero integers between-4 and + 4;is the set of all cations;is a corresponding cationiThe component ratio of (A);is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,3;is a power exponent ofWhen is a cationIs corresponding tomA descriptor; the performance parameters include the cation per unit of each crystaliBinding energy ofElastic tensor constant of each unit cell of cubic crystal containing cation AAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a).
3. The method of claim 2, wherein said per unit cation isiBinding energy ofThe calculation method comprises the following steps:
using the cell energy obtained in step 2-2Subtracting the sum of the energies of the single atoms of the same number and kind, and calculating by the formula:
4. The method according to claim 3, wherein the step 2-1 constructs an atomic structure model of the oxide crystal having different symmetries as a unit cell of molecular dynamics calculation, comprising:
(1) for cations with a valence of +1, according to CaF2The structure is crystal, oxygen ions form a face-centered cubic structure, and cations are in tetrahedral gaps of the face-centered cubic structure;
(2) for + 2-valent cations, crystals are constructed according to a NaCl structure, the cations form a face-centered cubic structure, and oxygen ions are in hexahedral gaps of the face-centered cubic structure;
(3) for cations with a valence of +3, according to α -Al2O3A structural structure crystal;
(4) for cations with a valence of +4, according to CaF2The structure makes crystal, the positive ions form a face-centered cubic structure, and the oxygen ions are in the tetrahedral gaps of the face-centered cubic structure.
5. The method according to claim 4, wherein the step 1 of constructing the elastic modulus database further comprises preprocessing the acquired elastic modulus data of oxide glass materials with different compositions:
judging whether the following two conditions are simultaneously satisfied or not according to the two glass components:
condition 1: in the glass components, the molar ratio difference value of the components of each oxide component is less than or equal to a first preset threshold, and the unit is percentage;
condition 2: the difference value of the elastic moduli of the two glasses is greater than a second preset threshold value, and the unit is percentage;
and if the two glass components are simultaneously satisfied, removing the corresponding two glass components and the corresponding elastic modulus data from the elastic modulus database.
6. The method according to claim 5, wherein the first predetermined threshold is 2% and the second predetermined threshold is 20%.
7. The method of claim 6, wherein the elastic network regression model predicts the elastic modulusyAnd descriptorsThe following relationship is satisfied:
wherein the content of the first and second substances,Perin order to be a set of performance parameters,is composed ofCorresponding regression coefficients, n being a non-zero integer from-4 to 4;
the loss function of the elastic network regression model is:
wherein the content of the first and second substances,for the loss function, N is the number of samples,in order to predict the value of the target,in the form of an actual value of the value,
8. An elastic modulus prediction system based on molecular dynamics and elastic network regression algorithm, the system comprising:
the elastic modulus database construction module is used for acquiring elastic modulus data of glass materials with different component compositions and constructing an elastic modulus database, wherein the elastic modulus database comprises glass components mapped one by one and elastic moduli corresponding to the glass components;
a descriptor construction module for constructing a descriptor containing a 'material gene' of the elastic modulus of the oxide glass material based on molecular dynamics;
the training data construction module is used for constructing a training set, a verification set and a test set on the basis of the elastic modulus database constructed by the elastic modulus database construction module and the descriptor constructed by the descriptor construction module;
the model construction training module is used for constructing an elastic modulus prediction model based on an elastic network regression model tree, and training the constructed elastic modulus prediction model according to a training set, a verification set and a test set constructed by the training data construction module to obtain a trained elastic modulus prediction model;
and the prediction module is used for constructing an elastic modulus prediction model trained by the training module by utilizing the model to predict the elastic modulus of the glass material to be predicted.
9. The elastic modulus prediction system of claim 8, wherein the descriptor construction module is configured to construct a descriptor containing a "material gene" of the elastic modulus of the oxide glass material, comprising
Constructing an atomic structure model of oxide crystals with different symmetries as a unit cell for molecular dynamics calculation;
for each cell structure obtained by construction, molecular dynamics calculation is carried out to obtain unit cell energy;
For each cell structure obtained by construction, obtaining a performance parameter set thereof through further calculation, and constructing a descriptor for machine learning:
In the formula (I), the compound is shown in the specification,nis a power-averaged exponent, and takes on all non-zero integers between-4 and + 4;is the set of all cations;is a corresponding cationiThe component ratio of (A);is one of a set of performance parameters,means when the power exponent isWhen is a cationIs corresponding toIndividual performance parameters;,3;is a power exponent ofWhen is a cationThe m-th descriptor corresponding to the structure of (1); the performance parameters include the cation per unit of each crystaliBinding energy ofEach of the cubic crystals containing the cation ASpring tensor constant of unit cellAndvolume of each unit cell of cubic crystals containing cation AAnd Voigt average shear modulusThe product of (a).
10. A computer device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the computer program is executed by the processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110715862.2A CN113255083B (en) | 2021-06-28 | 2021-06-28 | Elastic modulus prediction method based on molecular dynamics and elastic network regression model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110715862.2A CN113255083B (en) | 2021-06-28 | 2021-06-28 | Elastic modulus prediction method based on molecular dynamics and elastic network regression model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113255083A CN113255083A (en) | 2021-08-13 |
CN113255083B true CN113255083B (en) | 2021-10-01 |
Family
ID=77189814
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110715862.2A Active CN113255083B (en) | 2021-06-28 | 2021-06-28 | Elastic modulus prediction method based on molecular dynamics and elastic network regression model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113255083B (en) |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8820173B2 (en) * | 2009-03-06 | 2014-09-02 | Andrew C. Clark | Contact sensors and methods for making same |
CN111429980A (en) * | 2020-04-14 | 2020-07-17 | 北京迈高材云科技有限公司 | Automatic acquisition method for material crystal structure characteristics |
CN112037869A (en) * | 2020-09-01 | 2020-12-04 | 西南交通大学 | Method and system for testing material property and predicting strength, storage medium and computer equipment |
-
2021
- 2021-06-28 CN CN202110715862.2A patent/CN113255083B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN113255083A (en) | 2021-08-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103745273B (en) | Semiconductor fabrication process multi-performance prediction method | |
CN108268638A (en) | A kind of generation confrontation network distribution type implementation method based on Spark frames | |
CN111427750B (en) | GPU power consumption estimation method, system and medium of computer platform | |
CN111327046B (en) | Economic dispatching method for power system containing large-scale wind power scene | |
Hernández-Lobato et al. | Designing neural network hardware accelerators with decoupled objective evaluations | |
Hao et al. | A bi‐level ensemble learning approach to complex time series forecasting: Taking exchange rates as an example | |
CN102778555B (en) | Method for predicting concentration of gas dissolved in transformer oil | |
CN113255083B (en) | Elastic modulus prediction method based on molecular dynamics and elastic network regression model | |
Deng et al. | A hybrid aerodynamic optimization algorithm based on differential evolution and RBF response surface | |
CN109599866B (en) | Prediction-assisted power system state estimation method | |
Yamazaki | Efficient multi-objective shape optimization using proper orthogonal decomposition with variable fidelity concept | |
Nawel et al. | Modeling voltage real data set by a new version of Lindley distribution | |
CN114064746A (en) | Information equipment running state evaluation method | |
CN113312853B (en) | Density prediction method based on molecular dynamics and ridge regression algorithm | |
Arab et al. | Enhancing weighted uniform simulation for structural reliability analysis | |
Zhu et al. | Optimizing Shot Assignment in Variational Quantum Eigensolver Measurement | |
CN114925845B (en) | Machine learning construction method for embedding atomic potential function | |
Zhang et al. | A composite measurement scheme for efficient quantum observable estimation | |
Rochayani et al. | Finding biomarkers from a high-dimensional imbalanced dataset using the hybrid method of random undersampling and lasso | |
Porzelius et al. | The benefit of data-based model complexity selection via prediction error curves in time-to-event data | |
Lukaczyk | Surrogate modeling and active subspaces for efficient optimization of supersonic aircraft | |
Xue et al. | An Improved Algorithm Based on NSGA-II for Cloud PDTs Scheduling. | |
CN113076687B (en) | Electronic energy loss calculation method and device and electronic equipment | |
Helms et al. | Hybrid simulation of dynamic reaction networks in multi-level models | |
CN108875927A (en) | A kind of convergence method and device of high-dimensional deep learning model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |