Method for training artificial intelligence keying number prediction model and keying number prediction method

Abstract

The invention relates to a method for training an artificial intelligence keying number prediction model and a keying number prediction method, and the method for training the artificial intelligence keying number prediction model comprises the steps: obtaining noise data, inputting the noise data into a preset GAN generator, and obtaining a generated sample; setting a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models and a second number of dimethyl siloxane monomer structure models; applying a preset temperature, a preset pressure, the first number and the second number to a preset molecular dynamics simulation calculation process to construct a dimethyl siloxane dense membrane structure model, and obtaining the bond forming number of Si-O-Si bonds of the dimethyl siloxane dense membrane structure model; performing data enhancement based on the generative adversarial network to obtain multiple groups of generated data; and training a random forest model by using the generated data and a real sample to obtain the random forest model for predicting the bond forming number of Si-O-Si bonds of the dimethyl siloxane dense membrane structure model.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular, to a method for training an artificial intelligence bond number prediction model and a bond number prediction method. Background Art

[0002] Polydimethylsiloxane (PDMS) membrane, also called dimethylsiloxane membrane, exhibits excellent performance in gas adsorption separation due to its high gas flux, good chemical stability, flexibility and thermal stability. Its unique Si-O-Si bond structure and the hydrophobic methyl functional groups on its side chains give it excellent permeability, good hydrophobicity and organophilicity, allowing the PDMS membrane to preferentially permeate small organic molecules, effectively improving the selectivity in gas adsorption separation. Therefore, the number of Si-O-Si bonds is very likely to be the key parameter that determines the permeability characteristics of nanofiltration membranes.

[0003] However, there is a lack of systematic research on the bonding number of Si-O-Si bonds in polydimethylsiloxane in terms of the gas adsorption and separation performance of silicon-based nanofiltration membranes. This is mainly because the membranes studied are commercial products and the exact nature of their selective layers has not been fully revealed. Even so, the bonding number of Si-O-Si bonds in polydimethylsiloxane may be of great significance in gas adsorption and separation, but further in-depth research is still needed to systematically evaluate its specific impact. At present, the bonding number of Si-O-Si bonds is obtained by performing relevant statistics on the constructed dense membranes, and the construction of dense membranes requires a lot of computing power and time. Therefore, if the bonding number of Si-O-Si bonds of PDMS membranes can be quickly predicted, it can improve R&D efficiency and guide subsequent research. Summary of the invention

[0004] The purpose of the present disclosure is to provide a method for training an artificial intelligence bonding number prediction model and a bonding number prediction method, so as to train a model that can predict the bonding number of Si-O-Si bonds in a dimethylsiloxane dense film based on temperature, pressure, the number of siloxane prepolymer structure models and the dimethylsiloxane monomer structure models, thereby improving the scheduling speed of predicting the bonding number of its Si-O-Si bonds, reducing the consumption of resources such as computing power and time, and accelerating research and development efficiency.

[0005] In order to achieve the above-mentioned object, the first aspect of the present disclosure provides a method for training an artificial intelligence bonding number prediction model, the method comprising:

[0006] Obtain noise data, input the noise data into the preset GAN generator, and obtain generated samples;

[0007] Setting a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models;

[0008] Applying the preset temperature, the preset pressure, the first quantity and the second quantity to a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and obtaining the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model;

[0009] Repeat the following two steps:

[0010] The preset temperature, the preset pressure, the first quantity, the second quantity and the corresponding bond number form a real sample, the real sample and the generated sample are input into a preset GAN discriminator to obtain a first prediction result, a first loss function of the preset GAN discriminator is calculated according to the first prediction result, the parameters of the preset GAN generator are fixed, and the parameters of the preset GAN discriminator are updated according to the first loss function;

[0011] Input the generated sample into a preset GAN discriminator to obtain a second prediction result, calculate a second loss function of the preset GAN generator according to the second prediction result, fix the parameters of the preset GAN discriminator, and update the parameters of the preset GAN generator according to the second loss function;

[0012] When the two steps are repeated for a preset number of rounds, the two steps are stopped, and the noise data is input into the trained preset GAN generator to obtain multiple sets of generated data;

[0013] The generated data and the real samples are used to train the random forest model, and a random forest model is obtained that uses the temperature, pressure, the first quantity and the second quantity to predict the bonding number of Si-O-Si bonds in the dimethylsiloxane dense film structure model.

[0014] Optionally, the preset GAN generator is a fully connected neural network with 3 layers, the preset GAN discriminator is a fully connected neural network with 3 layers, and the activation functions of the preset GAN generator and the preset GAN discriminator are tanh activation functions.

[0015] Optionally, the first loss function is calculated based on the following formula:

[0016]

[0017] Among them, L D is the first loss function, x is the real sample, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E is the expectation, p data is the probability distribution of the real sample, p z is the probability distribution of the noise data.

[0018] Optionally, the second loss function is calculated based on the following formula:

[0019]

[0020] Among them, L G is the second loss function, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E is the expectation, p z is the probability distribution of the noise data.

[0021] Optionally, the preset number of rounds is 3000 rounds.

[0022] Optionally, applying the preset temperature, the preset pressure, the first quantity and the second quantity to a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and obtaining the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model includes:

[0023] Using Material Studio software to construct a siloxane prepolymer structure model and a dimethylsiloxane monomer structure model, and geometrically optimizing the siloxane prepolymer structure model and the dimethylsiloxane monomer structure model, respectively, and exporting pdb format files of the geometrically optimized siloxane prepolymer structure model and the dimethylsiloxane monomer structure model;

[0024] Opening a pdb format file in Packmol software, randomly mixing the first number of geometrically optimized siloxane prepolymer structure models and the second number of geometrically optimized dimethylsiloxane monomer structure models to establish a target polymer film initial structure model, and exporting a pdb format file of the target polymer film initial structure model;

[0025] The pdb format file of the initial structure model of the target polymer membrane is converted into an lmp format file using Atomsk software, and then the lmp file is converted into a data format file using OVITO software;

[0026] The data format file was opened using LAMMPS software. The initial structure model of the loaded target polymer membrane was statically minimized based on the conjugate gradient algorithm. Then the NVT ensemble was selected and the motion equation was integrated using the Verlet-Velocity algorithm. A temperature equilibrium of 50 ps was performed at the first preset temperature with a time step of 0.1 fs. The temperature was then adjusted using the Nose-Hoover algorithm. A 100 ps equilibrium simulation was performed at the second preset temperature with a standard relaxation time of 10 fs to obtain the initial dimethylsiloxane dense membrane structure model.

[0027] Select the NPT ensemble, set the pressure to the first preset pressure, the temperature to the third preset temperature, the simulation time to 1ns, and perform a volume compression simulation calculation based on molecular dynamics simulation on the dimethylsiloxane dense membrane structure model to obtain a dimethylsiloxane dense membrane structure model with a stable structure, and then count the number of Si-O-Si bonds in the dimethylsiloxane dense membrane structure model.

[0028] In a second aspect of the present disclosure, a bonding number prediction method is provided, the method comprising:

[0029] Execute the method described in the first aspect to obtain a trained random forest model for predicting the bonding number of Si-O-Si bonds in a dimethylsiloxane dense film structure model;

[0030] Obtaining a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models;

[0031] The preset temperature, the preset pressure, the first quantity and the second quantity are input into the random forest model to obtain the predicted number of Si-O-Si bonds of the dimethylsiloxane dense film structure model constructed under the conditions of the preset temperature, the preset pressure, the first quantity and the second quantity.

[0032] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of the method described in any one of the first aspect or the second aspect are implemented.

[0033] According to a fourth aspect of the present disclosure, there is provided an electronic device, comprising:

[0034] a memory having a computer program stored thereon;

[0035] A processor is used to execute the computer program in the memory to implement the steps of the method in any one of the first aspect or the second aspect.

[0036] Through the above technical solution, the preset GAN generator and preset GAN discriminator of the generative adversarial network are trained to enhance the data for constructing the dense membrane structure model of polydimethylsiloxane, increase its data volume, and generate more generated data that conforms to the distribution of real samples to overcome the challenge of data scarcity. Then, the generated data and real samples are used to train the random forest model. The rich data can improve the training effect, so that the trained random forest model can effectively capture the nonlinear relationship between variables and accurately predict the number of Si-O-Si bonds under different conditions, providing an efficient and low-cost solution for predicting the number of Si-O-Si bonds in the dense membrane structure model of dimethylsiloxane.

[0037] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0039] Figure 1 The present invention is a flowchart of a method for training an artificial intelligence bond number prediction model according to an exemplary embodiment.

[0040] Figure 2 is a flowchart of step S103 according to an exemplary embodiment.

[0041] Figure 3 The figure is a diagram of an initial structure model of a target polymer film according to an exemplary embodiment.

[0042] Figure 4 The figure is a model diagram of a dimethylsiloxane dense film structure according to an exemplary embodiment.

[0043] Figure 5 The figure is a flow chart of a method for predicting the number of bonds according to an exemplary embodiment.

[0044] Figure 6 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0045] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0046] Molecular dynamics (MD) is a technology that integrates multidisciplinary knowledge. It simulates the dynamic behavior of molecular systems based on the principles of Newtonian mechanics. By extracting samples of different states of the molecular system, it can approximate the configuration integral of the system to evaluate macroscopic aspects such as the thermodynamic properties of the system. Researchers can use it to explore the behavior of complex systems at the atomic and molecular scales, and it has a wide range of applications in many fields such as chemistry, physics, and materials science.

[0047] Although molecular dynamics (MD) simulations can deeply reveal the behavior of materials at the microscopic level, they are computationally expensive and the size of the data set is limited. Therefore, when faced with complex multivariable systems, researchers often face problems of inefficiency and insufficient data. Because the cross-linking process involves complex interactions of multiple variables, the efficiency of using classical methods to predict and optimize experimental conditions is limited. These complex nonlinear behaviors and interaction effects need to be gradually revealed and verified through a large amount of experimental work, which is costly and inefficient.

[0048] The rise of artificial intelligence (AI) and machine learning (ML) technologies has provided new possibilities for the optimization of such complex multivariable systems. ML algorithms are able to extract hidden patterns and laws from large amounts of experimental data without explicit programming. This allows researchers to predict and optimize material properties more efficiently with less experimental input. Especially in the cross-linking process of polymer materials, ML algorithms can help identify the changing laws of the number of bonds in important chemical bond structures in materials under different conditions, providing a scientific basis for material design. By combining molecular dynamics simulation (MD) and machine learning (ML) technology, researchers can theoretically simulate cross-linking behavior under different conditions, and conduct experimental verification on this basis, greatly reducing the number of experiments and reducing research costs.

[0049] Figure 1 is a flowchart of a method for training an artificial intelligence bond number prediction model according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0050] S101, obtain noise data, input the noise data into a preset GAN generator, and obtain a generated sample.

[0051] S102, setting a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models.

[0052] S103, applying the preset temperature, the preset pressure, the first quantity and the second quantity to a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and obtaining the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model.

[0053] Repeat the following two steps:

[0054] S104, the preset temperature, the preset pressure, the first quantity, the second quantity and the corresponding bond number are combined into a real sample, the real sample and the generated sample are input into the preset GAN discriminator together to obtain a first prediction result, a first loss function of the preset GAN discriminator is calculated according to the first prediction result, the parameters of the preset GAN generator are fixed, and the parameters of the preset GAN discriminator are updated according to the first loss function.

[0055] S105, input the generated sample into the preset GAN discriminator to obtain a second prediction result, calculate the second loss function of the preset GAN generator according to the second prediction result, fix the parameters of the preset GAN discriminator, and update the parameters of the preset GAN generator according to the second loss function.

[0056] S106, when the two steps are repeated for a preset number of rounds, the two steps are stopped and the noise data is input into the trained preset GAN generator to obtain multiple groups of generated data.

[0057] S107, using the generated data and the real samples to train the random forest model, and obtaining a random forest model that uses the temperature, pressure, the first quantity and the second quantity to predict the bonding number of Si-O-Si bonds in the dimethylsiloxane dense film structure model.

[0058] The preset GAN generator and the preset GAN discriminator together constitute a generative adversarial network (GAN) for data enhancement.

[0059] In step S101, random noise data conforming to Gaussian distribution may be obtained, and then the noise data may be input into a preset GAN generator, which may output a generated sample after calculation. The preset GAN generator may be a fully connected neural network with 3 layers, and the activation function may be a tanh activation function.

[0060] After step S101 is executed, step S102 is executed. In step S102, the preset temperature may include multiple sets of temperature data, and the preset pressure may include multiple sets of pressure data, which are used for setting corresponding conditions in the process of molecular dynamics simulation to construct a dimethylsiloxane (PDMS) dense film. The preset temperature, preset pressure, first number of siloxane prepolymer structure models, and second number of dimethylsiloxane monomer structure models for molecular dynamics simulation can be formulated according to actual conditions to be applied when performing molecular dynamics simulation.

[0061] After step S102 is executed, step S103 is executed. In step S103, a preset molecular dynamics simulation calculation process is applied to construct a PDMS dense film structure model. For example, a PDMS film structure can be constructed using chemical modeling software such as Material Studio, Packmol, and LAMMPS, and appropriate structural optimization is performed to obtain a stable film structure model. The structural optimization calculation used in the process is based on molecular dynamics calculation implementation, and some of the calculation processes need to be run under certain temperature and pressure conditions. Therefore, the temperature and pressure conditions in the preset molecular dynamics simulation calculation process are set to the corresponding preset temperature and preset pressure, and the first number of siloxane prepolymer structure models and the second number of dimethylsiloxane monomer structure models are used to construct the PDMS dense film structure model. After the PDMS dense film structure model is constructed, the bonding number of the Si-O-Si bond of the dense film structure model can be counted for subsequent machine learning model training. The Si-O-Si bond refers to the skeleton structure of the siloxane prepolymer in the dimethylsiloxane dense film, in which the silicon atoms are connected by oxygen bridge bonds. In this structure, the silicon atoms and oxygen atoms on the prepolymer backbone form chemical bonds with the silicon atoms and oxygen atoms at the ends of the dimethylsiloxane monomer, respectively, thereby constructing an overall cross-linked structure.

[0062] After step S103 is completed, step S104 can be executed. In step S104, the preset temperature, preset pressure, first quantity, second quantity and the corresponding bonding number of the PDMS dense film constructed using these conditions set in step S102 are used as real samples, and the real samples and the generated samples obtained in step S101 are input into the preset GAN discriminator as training data to obtain the first prediction result output by the preset GAN discriminator, for example, a number between 0 and 1 can be output to represent the probability of it being true, such as "1" indicates that the predicted sample is true, and "0" indicates that the predicted sample is false. After that, the parameters of the preset GAN generator are fixed, and the parameters of the preset GAN discriminator are trained and updated. Among them, the preset GAN discriminator can be a fully connected neural network with 3 layers, and its activation function is a tanh activation function.

[0063] The parameters of the preset GAN generator may be kept unchanged, and the first loss function of the preset GAN discriminator may be calculated according to the first prediction result. The first loss function may be calculated based on the following formula:

[0064]

[0065] Among them, L D is the first loss function, x is the real sample, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E represents the expectation, p data is the probability distribution of the real sample, pz is the probability distribution of the noise data.

[0066] Specifically, in the formula, G(z) represents the generated sample generated by the preset GAN generator based on the noise data, D(G(z)) can represent the first preset result of the preset GAN discriminator for the generated sample, and D(x) is the first preset result of the preset GAN discriminator for the real sample.

[0067] The first loss function is calculated according to the first prediction result, for example, the cross entropy loss function of the above formula is calculated, and then the parameters of the preset GAN discriminator can be updated using the gradient descent algorithm and the cross entropy loss function. After that, the generated sample and the real sample can be input into the preset GAN discriminator with updated parameters, and the above process is repeated to continue updating the parameters of the preset GAN discriminator until the preset condition is met, such as the absolute value of the first loss function is less than the threshold, or the parameters of the preset GAN discriminator are repeatedly updated a certain number of times.

[0068] After step S104 is completed, step S105 may be executed. In step S105, the generated sample is input into the preset GAN discriminator to obtain the second prediction result output by the preset GAN discriminator, and then the parameters of the preset GAN discriminator are kept unchanged, and the parameters of the preset GAN generator are trained and updated.

[0069] The parameters of the preset GAN discriminator may be kept unchanged, and the second loss function corresponding to the preset GAN generator may be calculated according to the second prediction result. The second loss function may be calculated based on the following formula:

[0070]

[0071] Among them, L G is the second loss function, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E is the expectation, p z is the probability distribution of the noise data.

[0072] Specifically, G(z) in the formula is the generated sample generated by the preset GAN generator based on the noise data, and D(G(z)) is the second prediction result of the preset GAN discriminator for the generated sample.

[0073] Calculate the second loss function according to the second prediction result, for example, calculate the above formula L GThe cross entropy loss function is then used to update the parameters of the preset GAN generator using the gradient descent algorithm and the cross entropy loss function. After that, the noise data can be input into the preset GAN generator after the parameter update to obtain a new generated sample, and then the new generated sample is input into the preset GAN discriminator, and the above process is repeated to continue updating the parameters of the preset GAN generator until the preset conditions are met, such as the second loss function value converges or the number of repeated updates of the preset GAN generator parameters reaches the preset number of rounds.

[0074] After step S105 is executed, it is possible to return to re-iterate step S104 and step S105. Executing step S104 and step S105 once can be regarded as one round. As the number of training rounds increases, the learning effect of the generative adversarial network composed of the preset GAN generator and the preset GAN discriminator will continue to improve, and the accuracy of data enhancement will also increase accordingly. Selecting a suitable N value is crucial to the result, where N represents the preset number of rounds for which steps S104 and S105 are repeatedly executed. N cannot be too large or too small. As N continues to increase, the generated samples generated by the preset GAN generator will become closer and closer to the real samples, but this may also lead to overfitting. On the other hand, when the preset number of rounds N is small, the data generated by the preset GAN generator is easily recognized by the preset GAN discriminator. Therefore, the appropriate N value must be carefully selected during the training process. When the preset number of rounds N = 3000, the optimal form of data enhancement can be achieved to cover the initial data distribution.

[0075] Error analysis of the generative adversarial network. Generally speaking, the errors of the preset GAN discriminator and the preset GAN generator are opposite. The better the preset GAN discriminator can distinguish true from false, the greater the error of the preset GAN generator. On the contrary, if the output of the preset GAN generator can make the preset GAN discriminator unable to distinguish, the error of the preset GAN generator is small, while the error of the preset GAN discriminator is large. After the training starts, the error changes of the preset GAN generator and the preset GAN discriminator are more stable. After a certain number of training rounds, the error converges. Based on this, it can also be determined that the optimal value of the preset number of rounds N in the parameter update process is 3000.

[0076] After step S105 is executed, step S106 can be executed. In step S106, if the repeated execution of steps S104 and S105 reaches a preset number of rounds, the repeated iteration of steps S104 and S105 can be stopped, and the noise data is input into the preset GAN generator after training to obtain multiple groups of generated data. At this time, the generated data generated by the preset GAN generator has a distribution close to that of the real sample, which is difficult to be identified by the preset GAN discriminator and can be used for subsequent random forest model training. In one possible implementation, after the generative adversarial network is trained, the samples of the training set can be expanded from hundreds to thousands. In the prediction of the bonding process of complex substances, other chemical bonds may be incorrectly predicted due to imperfect models or insufficient data. In order to improve the accuracy of the prediction, more diverse data can be generated by data enhancement technology (such as generative adversarial network GAN), thereby improving the accuracy of the prediction.

[0077] After step S106 is executed, step S107 can be executed. In step S107, the generated data and the real samples are used to form a training set, and the training set is used to train the random forest model. The trained random forest model can use temperature, pressure, the first quantity, and the second quantity to predict the number of Si-O-Si bonds in the PDMS membrane structure. The random forest model adopts a regression model, and the output value is continuous. 80% of the data in the training set is used to train the random forest model, and 20% of the data in the training set is used to test the training results. During training, the training set can be sampled multiple times, and each time the training sample is extracted from the training set using the bootstrap sampling method to obtain multiple groups of training samples, each group of training samples is a subset of the training set. For the random forest model, the number of decision trees can be set to a value between 50-100, the maximum number of features of each decision tree is a value between 2-4, and the minimum number of leaf nodes is 2. A decision tree is generated by each group of training samples to obtain multiple decision trees, and finally the output of each decision tree is averaged to obtain the prediction result of the final random forest model.

[0078] Through the above technical solution, the preset GAN generator and preset GAN discriminator of the generative adversarial network are trained to enhance the data for constructing the dense membrane structure model of polydimethylsiloxane, increase its data volume, and generate more generated data that conforms to the distribution of real samples to overcome the challenge of data scarcity. Then, the generated data and real samples are used to train the random forest model. The rich data can improve the training effect, so that the trained random forest model can effectively capture the nonlinear relationship between variables and accurately predict the number of Si-O-Si bonds under different conditions, providing an efficient and low-cost solution for predicting the number of Si-O-Si bonds in the dense membrane structure model of dimethylsiloxane.

[0079] Alternatively, see Figure 2 In step S103, the preset temperature, the preset pressure, the first quantity and the second quantity are applied to a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model is obtained, including:

[0080] S1031, using Material Studio software to construct a siloxane prepolymer structure model and a dimethylsiloxane monomer structure model, and geometrically optimize the siloxane prepolymer structure model and the dimethylsiloxane monomer structure model respectively, and export pdb format files of the geometrically optimized siloxane prepolymer structure model and the dimethylsiloxane monomer structure model.

[0081] S1032, opening a pdb format file in Packmol software, randomly mixing a first number of geometrically optimized siloxane prepolymer structure models and a second number of geometrically optimized dimethylsiloxane monomer structure models to establish an initial structure model of a target polymer film, and exporting a pdb format file of the initial structure model of the target polymer film.

[0082] S1033, converting the pdb format file of the target polymer film initial structure model into an lmp format file using Atomsk software, and then converting the lmp file into a data format file using OVITO software.

[0083] S1034, use LAMMPS software to open the data format file, perform static minimization on the loaded target polymer membrane initial structure model based on the conjugate gradient algorithm, then select the NVT ensemble, use the Verlet-Velocity algorithm to integrate the motion equation, perform 50ps temperature equilibrium at the first preset temperature, and the time step is 0.1fs. Then use the Nose-Hoover algorithm to adjust the temperature, and perform a 100ps equilibrium simulation at the second preset temperature with a standard relaxation time of 10fs to obtain the initial dimethylsiloxane dense membrane structure model.

[0084] S1035, select the NPT ensemble, set the pressure to the first preset pressure, the temperature to the third preset temperature, the simulation time to 1ns, perform a volume compression simulation calculation based on molecular dynamics simulation on the dimethylsiloxane dense membrane structure model, and obtain a dimethylsiloxane dense membrane structure model with a stable structure, and then count the bonding number of Si-O-Si bonds of the dimethylsiloxane dense membrane structure model.

[0085] In step S1031, a dimethylsiloxane monomer and tetraethyl silicate structure can be first established in Materials Studio software, four ethyl groups on tetraethyl silicate are removed, and two dimethylsiloxane monomers are connected to the silicon and oxygen atoms of the tetraethyl silicate monomer to form a siloxane prepolymer structure, wherein the molecular formula of dimethylsiloxane is -(Si(CH3)2-O-) n -, whose structural formula is as follows:

[0086]

[0087] The molecular formula of tetraethyl silicate monomer is Si(OEt)4(C8H 20 O4Si), whose structural formula is as follows:

[0088]

[0089] The COMPASSIII force field in the Forcite module of Materials studio software and the Atom-based and Eward methods can be used to calculate the van der Waals and electrostatic effects on the dimethylsiloxane monomer structure and the siloxane prepolymer structure respectively, and finally the geometric structure optimization of the constructed model can be achieved based on the principle of system energy minimization.

[0090] Specifically, you can use the Geometry optimization task in the Forcite module, using the Smart algorithm, with Energy of 0.001 kcal / mol and Force of Displacement The maximum number of iterations is 500 steps, the force field uses COMPASSIII force field, the charge is Forcefield assigned, the charge group is automatically assigned by the force field, the electrostatic term uses the Ewald summation method, the accuracy is 0.001 kcal / mol, the van der Waals term uses the Atom-based summation method, the truncation method uses the Cubic spline, and the truncation radius is Key width is Long range correction was used for geometry optimization during the calculation process. After the geometry optimization was completed, the pdb format files of the siloxane prepolymer structure model and the dimethylsiloxane monomer structure model were exported for subsequent calculations with other software.

[0091] After step S1031 is completed, step S1032 may be executed. In step S1032, a second number of geometrically optimized dimethylsiloxane monomer molecules and a first number of newly synthesized siloxane prepolymer molecules that are also geometrically optimized may be selected in the modeling software Packmol. Figure 3 , randomly placing these molecules in a In a cubic box, the amorphous mixed system under simulated actual conditions is used as the initial structure model of the target polymer film. Figure 3 The cube composed of the lines shown is a three-dimensional box, which is used to represent the core component of the crystal structure - the unit cell. It is the basic building block that reveals the orderly and periodic arrangement rules inside the crystal. This step allows dimethylsiloxane monomers and siloxane prepolymers to be randomly distributed in the same simulation environment, thereby simulating their interactions and structure formation processes. Through the random placement function of Packmol, the initial structure of the material can be accurately established, providing appropriate starting conditions for subsequent molecular dynamics simulations to further explore the changes in its structure and properties. After constructing the initial structure model of the target polymer film, export its pdb format file.

[0092] After step S1032 is completed, step S1033 can be executed. In step S1033, the pdb file of the initial structure model of the target polymer membrane is converted into the lmp format using the Atomsk software for further processing. Then, the lmp file is converted into a data file recognizable by LAMMPS using the OVITO software. In this way, a seamless transition from construction to simulation of the initial model is ensured, so that the molecular dynamics simulation can proceed smoothly, providing a reliable basis for subsequent simulation and analysis.

[0093] After step S1033 is completed, step S1034 is executed. In step S1034, an adapted in format file can be written in the program writing software according to the simulation format requirements, and integrated into the script to form an automatically executable file. This method not only simplifies the simulation process and improves efficiency, but also ensures that each step is automatically executed according to the preset parameters and conditions, thereby ensuring the accuracy and stability of the simulation.

[0094] In the script, the calculation process can be written, including the static energy minimization of the constructed initial amorphous structure model containing the second number of dimethylsiloxane monomer molecules and the first number of newly synthesized siloxane prepolymer molecules using the conjugate gradient algorithm (cg). This step involves fine-tuning the atomic coordinates of the initial model loaded in LAMMPS to find the most stable structure of the system. The conjugate gradient algorithm is an efficient global optimization technique suitable for unconstrained optimization problems. It is known for its iterative simplicity and low storage requirements. This optimization process is crucial to ensure the convergence and accuracy of subsequent molecular dynamics (MD) simulations.

[0095] Specifically, the LAMMPS software was used to analyze the stability and optimize the structure of the initial structure model of the target polymer membrane with the ReaxFF force field. The NVT ensemble was used in the simulation process, and the Verlet-Velocity algorithm was used to integrate the equation of motion with a time step of 0.1fs. The Nose-Hoover algorithm was used to adjust the temperature with a standard relaxation time of 10fs. First, a 50ps temperature equilibrium was performed at the first preset temperature with a time step of 0.1fs; then a 100ps equilibrium simulation was continued at the second preset temperature to further stabilize the system. During the simulation process, the system data file and trajectory file were output through the LAMMPS command to record and analyze the dynamic evolution of the system. The automated script was executed, and the new siloxane prepolymer structure was successfully constructed by repeatedly repeating the MD simulation at different temperatures under the NVT ensemble. These prepolymers were connected to the terminal silicon atoms and oxygen atoms in the dimethylsiloxane monomer structure through two oxygen active sites and two silicon active sites to form a stable three-dimensional network structure, and the initial dimethylsiloxane dense membrane structure model was obtained. The first preset temperature may be a temperature between 500K and 800K, and the second preset temperature may be a temperature between 298K and 330K.

[0096] After step S1034 is executed, step S1035 can be executed. In step S1035, Material Studio software can be used to set the pressure and temperature to the first preset pressure and the third preset temperature respectively under the NPT ensemble, and the simulation time is 1 ns. Then, the volume compression simulation calculation of the initial dimethylsiloxane dense membrane structure is performed to obtain the following: Figure 4 The dimethylsiloxane dense film structure model with a stable structure is shown, and the calculation process is realized based on molecular dynamics simulation calculation. Among them, the first preset pressure can be a pressure between 1kPa-100kPa, and the third preset temperature can be a temperature between 300K-800K.

[0097] The dimethylsiloxane dense film structure model constructed by the above scheme has good structural stability, so that under periodic boundary conditions, the structure of the model can be close to the actual physical state. This method optimizes the model through relaxation (energy optimization) and compression (adjusting the structural density) to make it as close as possible to the global energy minimum at the end of the simulation, thereby improving the accuracy of the simulation.

[0098] Figure 5 is a flow chart of a method for predicting the number of bonds according to an exemplary embodiment. Figure 5 As shown, the method includes:

[0099] S501, executing the method described in steps S101 to S107 to obtain a trained random forest model for predicting the bonding number of Si-O-Si bonds in a dimethylsiloxane dense film structure model.

[0100] S502, obtaining a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models.

[0101] S503, inputting the preset temperature, the preset pressure, the first quantity and the second quantity into the random forest model to obtain the predicted number of Si-O-Si bonds of the dimethylsiloxane dense membrane structure model constructed under the conditions of the preset temperature, the preset pressure, the first quantity and the second quantity.

[0102] In step S501, the method corresponding to steps S101 to S107 is executed to train a random forest model that can predict the bonding number of Si-O-Si bonds in the dimethylsiloxane dense film structure model using temperature, pressure, the first number and the second number.

[0103] After step S501 is completed, step S502 is executed. In step S502, one or more sets of calculation data may be set, each set of calculation data including a preset temperature, a preset pressure, a first quantity, and a second quantity as condition parameters for molecular dynamics simulation calculation to construct a PDMS dense membrane.

[0104] After step S502 is completed, step S503 is executed. In step S503, the preset temperature, the preset pressure, the first quantity and the second quantity are input into the random forest model to obtain the predicted output number of Si-O-Si bonds, which is the predicted number of bonds of a PDMS dense film constructed based on molecular dynamics simulation calculations assuming the input temperature, pressure and quantity condition parameters. The PDMS dense film is not actually constructed, which can save computing resources.

[0105] Through this technical solution, the trained random forest model can be used to efficiently estimate the number of Si-O-Si bonds in the dense membrane structure of dimethylsiloxane under various conditions, thereby eliminating the need to perform long molecular dynamics simulations each time. This method reduces the reliance on molecular dynamics simulations, saves computing resources and time costs, and is particularly suitable for situations where experimental parameters need to be changed frequently.

[0106] Figure 6 FIG. 1 is a block diagram of an electronic device according to an exemplary embodiment. Figure 5As shown, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0107] The processor 701 is used to control the overall operation of the electronic device 700 to complete the above-mentioned method for training an artificial intelligence bond number prediction model and all or part of the steps in the bond number prediction method. The memory 702 is used to store various types of data to support the operation of the electronic device 700, and these data may include, for example, instructions for any application or method used to operate on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 702 or sent through the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules, and the other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, is not limited here. Therefore, the corresponding communication component 705 may include: Wi-Fi module, Bluetooth module, NFC module, etc.

[0108] In an exemplary embodiment, the electronic device 700 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned method of training an artificial intelligence bond number prediction model and the bond number prediction method.

[0109] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the method for training an artificial intelligence bond number prediction model and the steps of the bond number prediction method are implemented. For example, the computer-readable storage medium may be the memory 702 including the program instructions, and the program instructions may be executed by the processor 701 of the electronic device 700 to complete the method for training an artificial intelligence bond number prediction model and the bond number prediction method.

[0110] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0111] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0112] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A method for training an artificial intelligence bond number prediction model, characterized in that: The method comprises: Obtain noise data, input the noise data into the preset GAN generator, and obtain generated samples; Setting a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models; Applying the preset temperature, the preset pressure, the first quantity and the second quantity to a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and obtaining the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model; Repeat the following two steps: The preset temperature, the preset pressure, the first quantity, the second quantity and the corresponding bond number form a real sample, the real sample and the generated sample are input into a preset GAN discriminator to obtain a first prediction result, a first loss function of the preset GAN discriminator is calculated according to the first prediction result, the parameters of the preset GAN generator are fixed, and the parameters of the preset GAN discriminator are updated according to the first loss function; Input the generated sample into a preset GAN discriminator to obtain a second prediction result, calculate a second loss function of the preset GAN generator according to the second prediction result, fix the parameters of the preset GAN discriminator, and update the parameters of the preset GAN generator according to the second loss function; When the two steps are repeated for a preset number of rounds, the two steps are stopped, and the noise data is input into the trained preset GAN generator to obtain multiple sets of generated data; The generated data and the real samples are used to train the random forest model, and a random forest model is obtained that uses the temperature, pressure, the first quantity and the second quantity to predict the bonding number of Si-O-Si bonds in the dimethylsiloxane dense film structure model.

2. The method according to claim 1, characterized in that The preset GAN generator is a fully connected neural network with 3 layers, the preset GAN discriminator is a fully connected neural network with 3 layers, and the activation functions of the preset GAN generator and the preset GAN discriminator are tanh activation functions.

3. The method according to claim 2, characterized in that The first loss function is calculated based on the following formula: Among them, L D is the first loss function, x is the real sample, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E is the expectation, p data is the probability distribution of the real sample, p z is the probability distribution of the noise data.

4. The method according to claim 2, characterized in that: The second loss function is calculated based on the following formula: Among them, L G is the second loss function, z is the noise data, D is the preset GAN discriminator, G is the preset GAN generator, E is the expectation, p z is the probability distribution of the noise data.

5. The method according to claim 2, characterized in that: The preset number of rounds is 3000 rounds.

6. The method according to claim 1, characterized in that The applying the preset temperature, the preset pressure, the first quantity and the second quantity in a preset molecular dynamics simulation calculation process to construct a dimethylsiloxane dense film structure model, and obtaining the bonding number of Si-O-Si bonds of the dimethylsiloxane dense film structure model, comprises: Using Material Studio software to construct a siloxane prepolymer structure model and a dimethylsiloxane monomer structure model, and geometrically optimizing the siloxane prepolymer structure model and the dimethylsiloxane monomer structure model, respectively, and exporting pdb format files of the geometrically optimized siloxane prepolymer structure model and the dimethylsiloxane monomer structure model; Opening a pdb format file in Packmol software, randomly mixing the first number of geometrically optimized siloxane prepolymer structure models and the second number of geometrically optimized dimethylsiloxane monomer structure models to establish a target polymer film initial structure model, and exporting a pdb format file of the target polymer film initial structure model; The pdb format file of the initial structure model of the target polymer membrane is converted into an lmp format file using Atomsk software, and then the lmp file is converted into a data format file using OVITO software; The data format file was opened using LAMMPS software. The initial structure model of the loaded target polymer membrane was statically minimized based on the conjugate gradient algorithm. Then the NVT ensemble was selected and the motion equation was integrated using the Verlet-Velocity algorithm. A temperature equilibrium of 50 ps was performed at the first preset temperature with a time step of 0.1 fs. The temperature was then adjusted using the Nose-Hoover algorithm. A 100 ps equilibrium simulation was performed at the second preset temperature with a standard relaxation time of 10 fs to obtain the initial dimethylsiloxane dense membrane structure model. Select the NPT ensemble, set the pressure to the first preset pressure, the temperature to the third preset temperature, the simulation time to 1ns, and perform a volume compression simulation calculation based on molecular dynamics simulation on the dimethylsiloxane dense membrane structure model to obtain a dimethylsiloxane dense membrane structure model with a stable structure, and then count the number of Si-O-Si bonds in the dimethylsiloxane dense membrane structure model.

7. A bonding number prediction method, characterized in that: The method comprises: Execute the method as claimed in claim 1 to obtain a trained random forest model for predicting the bonding number of Si-O-Si bonds in a dimethylsiloxane dense film structure model; Obtaining a preset temperature, a preset pressure, a first number of siloxane prepolymer structure models, and a second number of dimethylsiloxane monomer structure models; The preset temperature, the preset pressure, the first quantity and the second quantity are input into the random forest model to obtain the predicted number of Si-O-Si bonds of the dimethylsiloxane dense film structure model constructed under the conditions of the preset temperature, the preset pressure, the first quantity and the second quantity.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 7.

Images