A method for assessing explosive safety through neural network potential calculation

Abstract

The application provides a method for evaluating explosive safety by neural network potential calculation, comprising the following steps: S1, obtaining a microstructure image of a binder explosive by in-situ transmission electron microscopy (TEM); S2, establishing a molecular model of an energetic material and a molecular model of a binder according to the obtained microstructure image and modifying the microstructure image; S3, performing neural network training on data obtained by simulation, fitting a potential energy surface of the molecular model of the energetic material and the molecular model of the binder; S4, simulating evolution of the explosive with time under the action of a load; and S5, analyzing simulation results. The method for evaluating explosive safety by neural network potential calculation can evaluate all explosives that have been applied and have application potential by using the binder explosive real microstructure safety calculation method based on the neural network potential, greatly promotes application of new explosives, and brings great convenience to ammunition safety engineering personnel.

Description

Technical Field

[0001] This invention relates to the field of explosive safety, and in particular to a method for evaluating explosive safety through neural network potential calculation. Background Technology

[0002] To resolve the conflict between the power and safety of munitions, both domestic and international efforts are focused on developing novel explosives. Among these, polymer-bonded explosives, composed of a base explosive (such as RDX, HMX, and triaminotrinitrobenzene (TATB)) or mixtures thereof, combined with a polymer binder, possess excellent safety and high-energy properties, making them a crucial direction in current explosive formulation research. The impact initiation performance and detonation kinetics of explosives directly affect the performance indicators of weapon warheads. Studying the process and mechanism of explosive impact initiation and detonation growth is core to understanding the impact initiation performance, detonation performance, and safety performance of explosives, and forms the basis for the design of munition detonation power and safety. Furthermore, research on the hotspot formation mechanism and principles when an explosive is subjected to impact is extremely important.

[0003] Currently, macroscopic phenomenological reaction rate models are the most commonly used research method in China for explosive safety design. However, this method cannot adequately describe the mechanism of hotspot formation under impact. Establishing a detonation reaction flow model with a certain degree of universality to accurately describe the entire process of impact initiation and detonation growth of high-energy insensitive mixed explosives is the theoretical foundation for explosive power and safety design and is a current hot topic in detonation physics research both domestically and internationally. However, how to establish a model that can accurately describe the internal mesoscopic and microscopic structure of explosives and accurately assess explosive safety based on the complex atomic-scale physicochemical changes within the explosive is still a challenge. Existing methods cannot accurately assess the safety of explosives under extreme environments using atomic-scale models, and this problem remains to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a method for evaluating the safety of explosives through neural network potential calculation. The method for calculating the safety of the real microstructure of binder explosives based on neural network potential can evaluate all explosives currently in use and those with application potential, which will greatly promote the application of new explosives and bring great convenience to ammunition safety engineers.

[0005] To achieve the above objectives, the present invention provides a method for evaluating the safety of explosives through neural network potential calculation, comprising the following steps:

[0006] S1. Obtain microstructure images of binder explosives using in-situ transmission electron microscopy (TEM).

[0007] S2. Based on the obtained microstructure images, establish molecular models of energetic materials and binders, and modify the microstructure images.

[0008] S3. Based on density functional theory, quantum mechanical calculation methods are used to simulate the molecular models of energetic materials and binders. The simulation data is then used to train a neural network and fit the potential energy surfaces of the molecular models of energetic materials and binders.

[0009] S4. Apply external impact or high temperature, and use the established neural network potential to calculate and simulate the evolution of explosives over time under load.

[0010] S5. Analyze the simulation results, evaluate the temperature, stress, chemical and physical phenomena of the explosive under external load, and then evaluate the safety of the explosive.

[0011] Preferably, in step S1, the microstructure image includes the size of the energetic material crystals, internal cracks, pores, dislocations, and the size and distribution of the binder.

[0012] Preferably, in step S2, atoms are deleted and modified at the corresponding defect locations to conform to the microstructure image of a real binder explosive, thereby obtaining a binder explosive model containing the microstructure.

[0013] Preferably, in step S3, the basic idea of ​​density functional theory is to use electron density to describe the ground state energy of the system, and the energy function is expressed as:

[0014] E[ρ]=T s [ρ]+E ext [ρ]+E H [ρ]+E XC [ρ]+E elec [ρ];

[0015] Among them, T s [ρ] is the non-relativistic kinetic energy term; E ext [ρ] is the interaction energy between the electron and the external potential; E H [ρ] is the classical Coulomb interaction energy between electrons; E XC [ρ] is the exchange-correlation energy; E elec [ρ] is the energy of the interaction between electrons;

[0016] To solve for the electron density ρ(r), the Kohn-Sham equations need to be solved:

[0017]

[0018] Where, ψ i (r) is a Kohn-Sham orbital; ∈ i It is the energy of that orbit; V eff (r) is the effective potential, including external potential, Hartree potential, and exchange-correlation potential;

[0019] V eff (r)=V ext (r)+V H (r)+V XC (r);

[0020] Exchange-correlation energy E XC The calculation of [ρ] is crucial in density functional theory, specifically the exchange-correlation functional E. XC [ρ] is approximated in the following form:

[0021] E XC [ρ]=∫ρ(r)∈ XC (ρ(r))dr;

[0022] Where, ∈ XC It is a local density functional of the exchange-correlation energy;

[0023] The expression for the single-electron wavefunction:

[0024]

[0025] in, It is the kinetic energy term; It is an external potential; It is the Coulomb interaction between electrons; It is an exchange function;

[0026] The total energy of the system is the sum of the energies of the single-particle orbits:

[0027]

[0028] Where, ∈ i It is the energy of the electron orbit;

[0029] The energy calculation formula in quantum mechanics can be expressed by the following total energy term:

[0030] E total =E kinetic +E electrostatic +E exchange +E correlation ;

[0031] Among them, E kinetic It is the kinetic energy term; E electrostatic It is the Coulomb interaction energy; E exchange It is the energy exchange; E correlation It is related energy.

[0032] Preferably, in step S3, the theoretical method for the neural network potential is as follows:

[0033] The two-body embedded smooth version of the DP descriptor is a multi-body representation of the atomic local environment. The descriptor with complete information is given by the following formula:

[0034]

[0035] The number of adjacent atoms is less than N c Then the dimension is N c The matrix will be filled; It is a coordinate matrix, each row All can be constructed as:

[0036]

[0037] Where, r ij It is a relative coordinate, s(r) ij ) is defined as:

[0038]

[0039] in, From 1 to r s Switch to 0 at the cutoff radius; s(r) ij The function is smooth because the second derivative is continuous, and each row of the embedding matrix consists of nodes from the output layer of the NN function.

[0040]

[0041] Preferably, in step S5, the molecular temperature is calculated using the following formula:

[0042]

[0043] Where, k B This is the Boltzmann constant, taken as 1.380649 × 10⁻⁶. -23 m 2 kg / s 2 K, E kin N is the sum of the kinetic energies of all atoms in the system. DOF For the system as a whole, the degree of freedom;

[0044] Molecular stress is obtained by adding molecular kinetic energy and virial contribution, and is calculated using the following formula:

[0045] S ab =-mv a v b -W ab ;

[0046]

[0047] The first term of the virial contribution is the sum of the energy contributions of each pair of atoms; Np The first term represents the number of paired atoms; r1 and r2 are the positions of the two atoms in the pairwise interactions; F1 and F2 are the forces generated by the pairwise interactions; the second term is a bonding contribution of a similar form, N. b N represents the number of bonds formed; the last three terms are all similar contribution terms. a Contribution to key angles; N d Contribute to the dihedral angle first; N i The term represents the anomalous interaction term; the last term represents the KSpace contribution of the long-range Coulomb interaction.

[0048] Therefore, the present invention adopts the above-mentioned method for evaluating the safety of explosives by calculating the potential of a neural network. The method for calculating the safety of the real microstructure of binder explosives based on the neural network potential can evaluate all explosives that are currently in use and have application potential, which greatly promotes the application of new explosives and will bring great convenience to ammunition safety engineers. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the structure of a method for evaluating the safety of explosives by calculating the potential of a neural network according to the present invention;

[0050] Figure 2 This is a structural diagram of a floor suction head for a method of evaluating explosive safety through neural network potential calculation according to the present invention;

[0051] Figure 3 This is a schematic diagram of the controller connection for a method of evaluating explosive safety through neural network potential calculation according to the present invention. Detailed Implementation

[0052] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0054] Example 1

[0055] like Figures 1-3 As shown, this invention provides a method for evaluating the safety of explosives through neural network potential calculation, comprising the following steps:

[0056] S1. Obtain microstructure images of binder explosives using in-situ transmission electron microscopy (TEM).

[0057] Microstructure images include the crystal size of energetic materials, internal cracks, pores, dislocations, and the size and distribution of the binder;

[0058] S2. Based on the obtained microstructure images, establish molecular models of energetic materials and binders, and modify the microstructure images.

[0059] In step S2, atoms are deleted and modified at the corresponding defect locations to conform to the microstructure image of a real binder explosive, thereby obtaining a binder explosive model containing the microstructure.

[0060] S3. Based on density functional theory, quantum mechanical calculation methods are used to simulate the molecular models of energetic materials and binders. The simulation data is then used to train a neural network and fit the potential energy surfaces of the molecular models of energetic materials and binders.

[0061] In step S3, the basic idea of ​​density functional theory is to use electron density to describe the ground state energy of the system, and the energy function is expressed as:

[0062] E[ρ]=T s [ρ]+E ext [ρ]+E H [ρ]+E XC [ρ]+E elec [ρ];

[0063] Among them, T s [ρ] is the non-relativistic kinetic energy term (i.e., Kohn-Sham kinetic energy); E ext [ρ] is the interaction energy between the electron and the external potential; E H [ρ] is the classical Coulomb interaction energy (Hartree energy) between electrons; E XC [ρ] is the exchange-correlation energy (exchange-correlation functional); E elec [ρ] is the energy of the interaction between electrons;

[0064] To solve for the electron density ρ(r), the Kohn-Sham equations need to be solved:

[0065]

[0066] Where, ψ i (r) is a Kohn-Sham orbital; ∈ i It is the energy of that orbit; V eff (r) is the effective potential, including external potential, Hartree potential, and exchange-correlation potential;

[0067] V eff (r)=V ext (r)+V H (r)+V XC (r);

[0068] Exchange-correlation energy EXC The calculation of [ρ] is crucial in density functional theory, specifically the exchange-correlation functional E. XC [ρ] is approximated in the following form:

[0069] E XC [ρ]=∫ρ(r)∈ XC (ρ(r))dr;

[0070] Where, ∈ XC It is a local density functional of the exchange-correlation energy;

[0071] The expression for the single-electron wavefunction:

[0072]

[0073] in, It is the kinetic energy term; It is an external potential; It is the Coulomb interaction between electrons; It is an exchange function;

[0074] The total energy of the system is the sum of the energies of the single-particle orbits:

[0075]

[0076] Where, ∈ i It is the energy of the electron orbit;

[0077] The energy calculation formula in quantum mechanics can be expressed by the following total energy term:

[0078] E total =E kinetic +E electrostatic +E exchange +E correlation ;

[0079] Among them, E kinetic It is the kinetic energy term; E electrostatic It is the Coulomb interaction energy; E exchange It is the energy exchange; E correlation It is related energy.

[0080] In step S3, the theoretical method for the neural network potential is as follows:

[0081] The two-body embedded smooth version of the DP descriptor is a multi-body representation of the atomic local environment. The descriptor with complete information is given by the following formula:

[0082]

[0083] The number of adjacent atoms is less than N c Then the dimension is N c The matrix will be filled; It is a coordinate matrix, each row All can be constructed as:

[0084]

[0085] Where, r ij It is a relative coordinate, s(r) ij ) is defined as:

[0086]

[0087] in, From 1 to r s Switch to 0 at the cutoff radius; s(r) ij The function is smooth because the second derivative is continuous, and each row of the embedding matrix consists of nodes from the output layer of the NN function.

[0088]

[0089] S4. Apply external impact or high temperature, and use the established neural network potential to calculate and simulate the evolution of explosives over time under load.

[0090] S5. Analyze the simulation results, evaluate the temperature, stress, chemical and physical phenomena of the explosive under external load, and then evaluate the safety of the explosive.

[0091] In step S5, the molecular temperature is calculated using the following formula:

[0092]

[0093] Where, k B This is the Boltzmann constant, taken as 1.380649 × 10⁻⁶. -23 m 2 kg / s 2 K, E kin N is the sum of the kinetic energies of all atoms in the system. DOF The degree of freedom of the system as a whole.

[0094] Molecular stress is obtained by adding molecular kinetic energy and virial contribution, and is calculated using the following formula:

[0095] S ab =-mv a v b -W ab ;

[0096]

[0097] The first term of the virial contribution is the sum of the energy contributions of each pair of atoms; N pThe first term represents the number of paired atoms; r1 and r2 are the positions of the two atoms in the pairwise interactions; F1 and F2 are the forces generated by the pairwise interactions; the second term is a bonding contribution of a similar form, N. b N represents the number of bonds formed; the last three terms are all similar contribution terms. a Contribution to key angles; N d Contribute to the dihedral angle first; N i The term represents the anomalous interaction term; the last term represents the KSpace contribution of the long-range Coulomb interaction.

[0098] Therefore, the present invention adopts the above-mentioned method for evaluating the safety of explosives by calculating the potential of a neural network. The method for calculating the safety of the real microstructure of binder explosives based on the neural network potential can evaluate all explosives that are currently in use and have application potential, which greatly promotes the application of new explosives and will bring great convenience to ammunition safety engineers.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for evaluating the safety of explosives using neural network potential calculation, characterized in that, Includes the following steps: S1. Obtain microstructure images of binder explosives using in-situ transmission electron microscopy (TEM). In step S1, the microstructure image includes the size of the energetic material crystals, internal cracks, pores, dislocations, and the size and distribution of the binder; S2. Based on the obtained microstructure images, establish molecular models of energetic materials and binders, and modify the microstructure images. S3. Based on density functional theory, quantum mechanical calculation methods are used to simulate the molecular models of energetic materials and binders. The simulation data is then used to train a neural network and fit the potential energy surfaces of the molecular models of energetic materials and binders. In step S3, the basic idea of ​​density functional theory is to use electron density to describe the ground state energy of the system, and the energy function is expressed as: ； in, It is a non-relativistic kinetic energy term; It is the interaction energy between electrons and external potential; It is the classical Coulomb interaction energy between electrons; It is exchange-correlation energy; It is the energy of the interaction between electrons; To solve for the electron density The Kohn-Sham equations need to be solved: ； in, It is the Kohn-Sham orbit; It is the energy of that orbit; It is an effective potential, including external potential, Hartree potential, and exchange-correlation potential; ； Exchange-correlation energy The calculation of is crucial in density functional theory, specifically in exchange-correlation functionals. It is approximated as follows: ； in, It is a local density functional of the exchange-correlation energy; The expression for the single-electron wavefunction: ； in, It is the kinetic energy term; It is an external potential; It is the Coulomb interaction between electrons; It is an exchange function; The total energy of the system is the sum of the energies of the single-particle orbits: ； in, It is the energy of the electron orbital; The energy calculation formula in quantum mechanics can be expressed by the following total energy term: ； in, It is the kinetic energy term; It is the Coulomb interaction energy; It is energy exchange; It is related energy; S4. Apply external impact or high temperature, and use the established neural network potential to calculate and simulate the evolution of explosives over time under load. In step S4, the theoretical method for the neural network potential is as follows: The two-body embedded smooth version of the DP descriptor is a multi-body representation of the atomic local environment. The descriptor with complete information is given by the following formula: ； The number of adjacent atoms is less than The dimension is The matrix will be filled; It is a coordinate matrix, each row All can be constructed as: ； in, These are relative coordinates; Defined as: ； in, From 1 Switch to 0 at the cutoff radius; The function is smooth because the second derivative is continuous, and each row of the embedding matrix consists of nodes from the output layer of the NN function: ； S5. Analyze the simulation results, evaluate the temperature, stress, chemical and physical phenomena of the explosive under external load, and then evaluate the safety of the explosive.

2. The method for evaluating the safety of explosives by calculating potential using a neural network according to claim 1, characterized in that, In step S2, atoms are deleted and modified at the corresponding defect locations to conform to the microstructure image of a real binder explosive, thereby obtaining a binder explosive model containing the microstructure.

3. The method for evaluating the safety of explosives by calculating potential using a neural network according to claim 1, characterized in that, In step S5, the molecular temperature is calculated using the following formula: ； in, It is the Boltzmann constant, taken as ; It is the sum of the kinetic energies of all atoms in the system. For the system as a whole, the degree of freedom; Molecular stress is obtained by adding molecular kinetic energy and virial contribution, and is calculated using the following formula: ； ； The first of the virial contributions is the sum of the energy contributions of each pair of atoms; The number of paired atoms; and It refers to the positions of two atoms in a pairwise interaction; and The first term is the force generated by pairwise interactions; the second term is a bonding contribution of a similar form. This represents the number of bonds formed; the last three terms are all similar contribution terms. Contribution to key corners; Contribute first to the dihedral angle; The term represents the anomalous interaction term; the last term represents the KSpace contribution of the long-range Coulomb interaction.

Images