PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
February 28, 2026Chinese Science Bulletin (Chinese Version)0 citationsOpen Access

FIND Multiphysics: A High-Performance Intelligent Simulation Platform for Multiphysics Coupling

JYJia YiYHYinan HanJWJinlin Wang

Key Points

  • The aim is to develop a high-performance simulation platform for multiphysics coupling to optimize design and costs.
  • Development of the FIND Multiphysics platform using modular and scalable strategies.
  • Integration of finite element, finite volume, and finite difference numerical methods.
  • Combination of traditional numerical methods with deep learning for solving differential equations.
  • Successfully deployed FIND Multiphysics V1.0 on Tianhe-2 supercomputer.
  • Capable of handling 100,000 cores for forward problems and 1 million cores for inverse problems.
  • Validated on industrial-grade grids, showing reliable performance and promising application prospects.

Abstract

多物理场耦合仿真通过计算机模拟多个物理现象及其相互耦合作用的过程,广泛应用于航空航天、能源化工、高端装备与工业制造等复杂系统的仿真分析与优化控制。其目标在于实现设计、性能与成本的协同优化,并通过故障预测提升风险管控能力,逐步发展为面向未来复杂工业场景的通用系统级工具。现有多物理场仿真软件普遍存在可扩展性有限、模块化与智能化不足,以及在大规模复杂仿真中精度与性能欠佳等问题。为了应对这些挑战,本文提出了高度模块化、可扩展和用户友好的3M策略,并在此基础上研发了灵活、智能且高性能的多物理场仿真软件平台——FIND Multiphysics。该平台以有限元、有限体积、有限差分等传统数值方法求解器为基础底座,同时前瞻性地融合了物理信息与数据驱动求解的两大技术路径,构建了传统数值方法与深度学习方法双驱的先进偏微分方程正问题与反问题求解体系。该软件平台可灵活集成大语言模型等前沿技术,从而增强仿真能力并降低操作门槛。目前,我们开发的FIND Multiphysics V1.0已部署在国家超级计算广州中心的“天河二号”星逸平台上,专为高性能计算环境而设计,可稳定支撑十万核规模的正问题求解与百万核规模的反问题求解,该方案已通过亿级网格的工业级验证,性能可靠,应用前景广阔。

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Yi et al. (2026) studied this question.

synapsesocial.com/papers/69a285da0a974eb0d3c00d6chttps://doi.org/10.1360/csb-2025-5886
Ask AI
Helpful
Bookmark
Share
View Full Paper