火炸药学报

针对一种四组元HTPB推进剂(HTPB/Al/AP/RDX)关键组分,基于第一性原理计算的数据集,使用深度神经网络模型开发了一个机器学习势函数; 基于新开发的势函数,建立了四组元HTPB推进剂燃面模型,并进行了大规模的分子动力学模拟计算,对推进剂燃烧时的微观结构、温度、反应组分的时空演化进行了系统统计分析。结果表明,新开发的势函数能够准确描述推进剂组分单质及两两之间界面的能量和受力特性,是一个高精度、高效率的机器学习势函数; 燃面模型实现了对推进剂燃烧过程中的AP、RDX、HTPB热解过程精准模拟,阐明了扩散火焰的形成机理以及铝粉从燃面剥离等微观过程,揭示了其中的各组分界面相互作用机制。表明分子动力学模拟能够在原子尺度上实现时间分辨的三维重建,进而获得推进剂燃烧的微观机理,可为固体推进剂的理论研究提供了新的工具。

A machine learning potential function was developed using a deep neural network model based on a first principles calculation dataset for the key component of a four component HTPB propellant(HTPB/Al/AP/RDX). Based on the newly developed potential function, a four component HTPB propellant combustion surface model was established, and a large-scale molecular dynamics simulation was conducted to systematically analyze the spatiotemporal evolution of microstructure, temperature, and reaction components during propellant combustion. The results show that the newly developed potential function can accurately describe the energy and force characteristics of the propellant components and the interface between them, and is a high-precision and high-efficiency machine learning potential function; The combustion surface model accurately simulates the pyrolysis process of AP, RDX, and HTPB during propellant combustion, elucidates the formation mechanism of diffusion flames and the microscopic process of aluminum powder peeling off from the combustion surface, and reveals the interaction mechanism of each component interface. This indicates that molecular dynamics simulation can achieve time-resolved three-dimensional reconstruction at the atomic scale, thereby obtaining the microscopic mechanism of propellant combustion, providing a new tool for the theoretical research of solid propellants.

引言
1 模拟方法
2 结果与讨论
3 结论

图1 机器学习势函数开发流程<br/>Fig.1 Development process of machine learning potential function

图1 机器学习势函数开发流程
Fig.1 Development process of machine learning potential function

图2 HTPB/Al/AP/RDX四组元推进剂装填模型(图中的 C/H/N/O/Al/Cl 分别使用灰/白/蓝/红/银/绿色表示)<br/>Fig.2 HTPB/Al/AP/RDX four-component propellant model (The C/H/N/O/Al/Cl in the diagram are represented in gray/white/blue/red/silver/green respectively)

图2 HTPB/Al/AP/RDX四组元推进剂装填模型(图中的 C/H/N/O/Al/Cl 分别使用灰/白/蓝/红/银/绿色表示)
Fig.2 HTPB/Al/AP/RDX four-component propellant model (The C/H/N/O/Al/Cl in the diagram are represented in gray/white/blue/red/silver/green respectively)

图3 单组分和两两界面的机器学习势函数的能量与受力精度测试<br/>Fig.3 Energy and force accuracy testing of machine learning potential for single-component and interfaces

图3 单组分和两两界面的机器学习势函数的能量与受力精度测试
Fig.3 Energy and force accuracy testing of machine learning potential for single-component and interfaces

表1 机器学习势函数对推进剂单质和界面组分的能量与受力精度<br/>Table 1 Prediction accuracy of the machine learning potential function for energies and forces on single and interfacial components of solid propellants

表1 机器学习势函数对推进剂单质和界面组分的能量与受力精度
Table 1 Prediction accuracy of the machine learning potential function for energies and forces on single and interfacial components of solid propellants

图4 推进剂燃烧过程中结构演化<br/>Fig.4 Structural evolution during propellant combustion

图4 推进剂燃烧过程中结构演化
Fig.4 Structural evolution during propellant combustion

图5 推进剂扩散火焰微结构局部放大图(t=34ps)<br/>Fig.5 Local magnification of diffusion flame microstructure(t=34ps)

图5 推进剂扩散火焰微结构局部放大图(t=34ps)
Fig.5 Local magnification of diffusion flame microstructure(t=34ps)

图6 推进剂燃烧过程中温度演化<br/>Fig.6 Temperature evolution during propellant combustion

图6 推进剂燃烧过程中温度演化
Fig.6 Temperature evolution during propellant combustion

图7 AP颗粒主要分解产物HClO4组分的时空演化<br/>Fig.7 The spatiotemporal evolution of HClO4 which is the main decomposition product in AP particles

图7 AP颗粒主要分解产物HClO4组分的时空演化
Fig.7 The spatiotemporal evolution of HClO4 which is the main decomposition product in AP particles

图8 O2/Cl2组分演化(图中浓度为两组分总浓度)<br/>Fig.8 Evolution of O2/Cl2 components(The concentration in the figure represents the total concentration of two components)

图8 O2/Cl2组分演化(图中浓度为两组分总浓度)
Fig.8 Evolution of O2/Cl2 components(The concentration in the figure represents the total concentration of two components)

图9 分解产物 NO/NO2分子的时空演化(图中浓度为两组分总浓度)<br/>Fig.9 The spatiotemporal evolution of NO/NO2 molecules(The concentration in the figure represents the total concentration of two components)

图9 分解产物 NO/NO2分子的时空演化(图中浓度为两组分总浓度)
Fig.9 The spatiotemporal evolution of NO/NO2 molecules(The concentration in the figure represents the total concentration of two components)

图10 CO/CO2组分演化(图中浓度为两组分总浓度)<br/>Fig.10 Evolution of CO/CO2 components(The concentration in the figure represents the total concentration of two components)

图10 CO/CO2组分演化(图中浓度为两组分总浓度)
Fig.10 Evolution of CO/CO2 components(The concentration in the figure represents the total concentration of two components)

图11 H2O组分演化<br/>Fig.11 Evolution of H2O components

图11 H2O组分演化
Fig.11 Evolution of H2O components

图12 Al组分演化过程<br/>Fig.12 Evolution of Al components

图12 Al组分演化过程
Fig.12 Evolution of Al components

[1] YADAV N, SRIVASTAVA P K, VARMA M. Recent advances in catalytic combustion of AP-based composite solid propellants[J]. Defence Technology, 2021, 17(3): 1013-1031.
[2]王道林, 钱勖, 孙晓飞, 等. 复合固体推进剂力学性能调控方法研究进展[J]. 固体火箭技术, 2022, 45(5): 729-740.
[3]BECKSTEAD M. Solid propellant combustion mechanisms and flame structure[J]. Pure and Applied Chemistry, 1993, 65(2): 297-307.
[4]BECKSTEAD M W, DERR R, PRICE C. A model of composite solid-propellant combustion based on multiple flames[J]. AIAA Journal, 1970, 8(12): 2200-2207.
[5]颜密. AP/HTPB复合推进剂压力耦合响应测试方法与数值模型研究[D]. 北京: 北京理工大学, 2017.
[6]张晏榕, 仓宇, 杨硕, 等. 非均质推进剂燃烧演化模拟:算法与验证[J]. 火炸药学报, 2023,46(6): 553-560.
[7]黄蒙, 丁黎, 常海, 等. HTPB/Al/AP/RDX复合推进剂组元之间相互作用研究[J]. 火炸药学报, 2020, 43(2): 203-207.
[8]吴迎春, 卓著, 吴世曦, 等. 基于高速离轴全息可视化方法的Al/AP/HTPB推进剂燃烧性能研究[J]. 火炸药学报, 2023, 46(2): 179-184.
[9]MARTIN J W, SALAMANCA M, KRAFT M. Soot inception: Carbonaceous nanoparticle formation in flames[J/OL]. Progress in Energy and Combustion Science, 2022, 88: 100956.
[10]MARTIN J W, PASCAZIO L, MENON A, et al. π-Diradical aromatic soot precursors in flames[J]. Journal of the American Chemical Society, 2021, 143(31): 12212-12219.
[11]CHU Q, CHANG X, CHEN D. A physiochemical model for the combustion of aluminum nano-agglomerates in high-speed flows[J/OL]. Combustion and Flame, 2021, 237: 111739.
[12]KANG J, HAN B. First-principles characterization of the unknown crystal structure and ionic conductivity of Li7P2S8I as a solid electrolyte for high-voltage Li ion batteries[J]. The Journal of Physical Chemistry Letters, 2016, 7(14): 2671-2675.
[13]YEH I C, ANDZELM J W. Computational study of structural and energetic properties of ammonium perchlorate at interfaces[J/OL]. The Journal of Physical Chemistry C, 2021, 125(22): 12297-12304.
[14]VAN DUIN A C T, DASGUPTA S, LORANT F, et al. ReaxFF: A reactive force field for hydrocarbons[J/OL]. The Journal of Physical Chemistry A, 2001, 105(41): 9396-9409.
[15]WANG N, PENG J, PANG A, et al. Thermodynamic simulation of the RDX-aluminum interface using ReaxFF molecular dynamics[J/OL]. The Journal of Physical Chemistry C, 2017, 121(27): 14597-14610.
[16]CHU Q, WEN M, FU X, et al. Reaction network of ammonium perchlorate(AP)decomposition: The missing piece from atomic simulations[J/OL]. The Journal of Physical Chemistry C, 2023, 127(27): 12976-12982.
[17]WANG H, ZHANG L, HAN J. DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics[J]. Computer Physics Communications, 2018, 228: 178-184.
[18]KÜHNE T D, IANNUZZI M, DEL BEN M, et al. CP2K: An electronic structure and molecular dynamics software package-quickstep: Efficient and accurate electronic structure calculations[J]. Journal of Chemical Physics, 2020, 152(19): 194103.
[19]CHANG X, CHU Q, CHEN D. Monitoring the melting behavior of boron nanoparticles using a neural network potential[J/OL]. Physical Chemistry Chemical Physics, 2023, 25(18): 12841-12853.
[20]CHU Q, LUO K H, CHEN D. Exploring complex reaction networks using neural network-based molecular dynamics simulation[J]. Journal of Physical Chemistry Letters, 2022, 13: 4052-4057.
[21]CHU Q, CHANG X, MA K, et al. Revealing the thermal decomposition mechanism of RDX crystals by a neural network potential[J/OL]. Physical Chemistry Chemical Physics, 2022, 24(42): 25885-25894.
[22]THOMPSON A P, AKTULGA H M, BERGER R, et al. LAMMPS-a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales[J/OL]. Computer Physics Communications, 2022, 271: 108171.
[23]ZENG J, CAO L, CHIN C H, et al. ReacNetGenerator: an automatic reaction network generator for reactive molecular dynamics simulations[J/OL]. Physical Chemistry Chemical Physics, 2020, 22(2): 683-691.

备注

引言

1 模拟方法

2 结果与讨论

3 结论

期刊信息

备注

引言

1 模拟方法

2 结果与讨论

3 结 论

期刊信息

3 结论