计算机科学 ›› 2024, Vol. 51 ›› Issue (9): 23-30.doi: 10.11896/jsjkx.230400010
刘人僪1,2, 陈欣3, 商红慧2, 张云泉2
LIU Renyu1,2, CHEN Xin3, SHANG Honghui2, ZHANG Yunquan2
摘要： 核反应堆压力容器是核电站中最重要的部件之一,在使用过程中通常会受到辐照损伤,这极大影响了其使用寿命,给核电站的安全带来隐患。原子动力学蒙特卡洛方法(AKMC)是研究材料辐照损伤的有效理论方法,可以与计算机数值模拟进行结合来研究压力容器的微结构演变。辐照损伤的材料存在缺陷,原子间相互作用建模时需要考虑非球对称相互作用,但TensorKMC在计算时并没有考虑到原子的角向作用。文中针对该问题,提出了一种包含角向相互作用、可以与TensorKMC的三重编码完美结合的指纹建模方法,并可利用多重度对角向指纹的计算过程进行化简。文中在TensorKMC程序中实现了该方法,测试结果显示角向指纹对势函数的精度有显著影响,最大角动量越高,势函数越精准,程序的模拟耗时也会显著增加。同时,也针对TensorKMC的原子势函数的激活函数开展了测试,结果表明梯度光滑的Softplus和SquarePlus相比初版TensorKMC所用的ReLU在拟合高维势能面时有明显的优势,在最大角动量较低时ReLU有性能优势,但随着最大角动量的增大,不同激活函数对总体模拟时间几乎无特别影响。因此,在实际研究中推荐使用梯度光滑的激活函数。
中图分类号:
| [1]GARCIA C C,WAGNER G J,TIKARE V,et al.Crossing the mesoscale no-mans land via parallel kinetic Monte Carlo[R].Sandia National Laboratories(SNL),Albuquerque,NM,and Livermore,CA(United States),2009. [2]BECQUART C S,DOMAIN C.Introducing chemistry in ato-mistic kinetic Monte Carlo simulations of Fe alloysunder irradia-tion[J].Physica Status Solidi(B),2010,247(1):9-22. [3]MARTIN-BRAGADO I,RIVERA A,VALLES G,et al.MMonCa:An Object Kinetic Monte Carlo simulator for damage irradia-tion evolution and defect diffusion[J].Computer Physics Communications,2013,184(12):2703-2710. [4]JOURDAN T,BOCQUET J L,SOISSON F.Modeling homogeneous precipitationwith an event-based Monte Carlo method:Application to the case of Fe-Cu[J].Acta materialia,2010,58(9):3295-3302. [5]BURGER G B,GUPTA A K,JEFFREY P W,et al.Microstructural control of aluminum sheet used in automotive applications[J].Materials Characterization,1995,35(1):23-39. [6]GUO H,ENOMOTO M,SHANG C J.Simulation of bcc-Cuprecipitation in ternary Fe-Cu-M alloys[J].Computational Materials Science,2018,141:101-113. [7]LIU H,WANG X,CUI W,et al.The effect of dislocations on irradiation-induced vacancy-like defects in FeCu alloy and reactor pressure vessel steel[J].Journal of Nuclear Materials,2019,524:80-89. [8]HUANG S,ZHANG C H,SUN J,et al.Formatin and Migration Mechanism of the Vacancy in Three Typical Structures Metal[J].Applied Physics,2012,2(2):50-54. [9]WANG C Y,WANG Z Q,MENG Q Y.Comparative study of the first-principles and empirical potential simulation of vacancies in silicon[J].Acta Physica Sinica,2010(5):3370-3376. [10]MARTYNEC T,KARAPANAGIOTIS C,KLAPP S H L,et al.Machine learning predictions of surface migration barriers in nucleation and non-equilibrium growth[J].Communications Materials,2021,2(1):1-9. [11]XU B,ZHANG J,MA S,et al.Revealing the crucial role of rough energy landscape on self-diffusion in high-entropy alloys based on machine learning and kinetic Monte Carlo[J].Acta Materialia,2022(234):118051-118062. [12]ZHANG S,ZHANG P,CHERN G W.Anomalous phase separation in a correlated electron system:Machine-learning-enabled large-scale kinetic Monte Carlo simulations[J].Proceedings of the National Academy of Sciences,2022,119(18):e2119957119. [13]SHANG H,CHEN X,GAO X,et al.TensorKMC:Kineticmonte carlo simulation of 50 trillion atoms driven by deep lear-ning on a new generation of sunway supercomputer[C]//Proceedings of the International Conference for High Perfor-mance Computing,Networking,Storage and Analysis.2021:1-14. [14]CASTIN N,PASCUET M I,MESSINA L,et al.Advancedatomistic models for radiation damage in Fe-based alloys:Contributions and future perspectives from artificial neural networks[J].Computational Materials Science,2018,148:116-130. [15]CASTIN N,MESSINA L,DOMAIN C,et al.Improved atomistic Monte Carlo models based on ab-initio-trained neural networks:Application to FeCu and FeCr alloys[J].Physical Review B,2017,95(21):214117. [16]MISHIN Y,LOZOVOI A Y.Angular-dependent interatomic po-tential for tantalum[J].Acta Materialia,2006,54(19):5013-5026. [17]SOISSON F,BECQUART C S,CASTIN N,et al.Atomistic Kinetic Monte Carlo studies of microchemical evolutions driven by diffusion processes under irradiation[J].Journal of Nuclear Materials,2010,406(1):55-67. [18]BEHLER J,PARRINELLO M.Generalized neural-network representation of high-dimensional potential-energy surfaces[J].Physical Review Letters,2007,98(14):146401. [19]VALLE M,OGANOV A R.Crystal structures classifier for an evolutionary algorithm structure predictor[C]//2008 IEEE Symposium on Visual Analytics Science and Technology.IEEE,2008:11-18. [20]DAW M S.Model of metallic cohesion:The embedded-atommethod[J].Physical Review B,1989,39(11):7441-7452. [21]BASKES M I,JOHNSON R A.Modified embedded atom potentials for HCP metals[J].Modelling and Simulation in Materials Science and Engineering,1994,2(1):147-163. [22]LENOSKY T J,SADIGH B,ALONSO E,et al.Highly opti-mized empirical potential model of silicon[J].Modelling and Simulation in Materials Science and Engineering,2000,8(6):825-841. [23]BASKES M I.Modified embedded-atom potentials for cubic materials and impurities[J].Physical review B,1992,46(5):2727. [24]BASKES M I.Atomistic model of plutonium[J].Physical Review B,2000,62(23):15532. [25]CHEN X,GAO X Y,ZHAO Y F,et al.TensorAlloy:An automatic atomistic neural network program for alloys[J].Computer Physics Communications,2020,250:107057. [26]LI Y B,GUO P Y,ZHANG S Y.Research on Activation Function in Deep Convolutional Neural Network[J].Computer Technology and Development,2021,31(9):61-66. [27]ZHAO H,LIU F,LI L,et al.A novel softplus linear unit for deep convolutional neural networks[J].Applied Intelligence,2018,48:1707-1720. [28]BARRON J T.Squareplus:A softplus-like algebraic rectifier[J].arXiv:2112.11687,2021. [29]VINCENT E,BECQUART C S,DOMAIN C.Solute interaction with point defects in α Fe during thermal ageing:A combined ab initio and atomic kinetic Monte Carlo approach[J].Journal of Nuclear Materials,2006,351(1／2／3):88-99. |
|
||
首页