Beihang University
We conduct research on multiscale gas-flow modeling and computation that bridges the continuum and rarefied regimes. Our work spans kinetic/particle methods and hybrid solvers; high-enthalpy flows and associated high-temperature gas effects; high-performance computing; and data-driven approaches for model discovery, equation solving, and aerodynamic design—including symbolic regression, PINNs, surrogate modeling, and reinforcement-learning–based optimization.
- Leader: Prof. Jun Zhang
- Google Scholar —Jun Zhang (张俊)
- Modeling and simulation of multiscale nonequilibrium gas flows
- Kinetic theory, stochastic particle methods, and hybrid solvers
- Open-source flow solver: Adaptive meshing/time-stepping and large-scale parallel computing
- Data-driven modeling: Symbolic regression, governing equation discovery, and gene expression programming
- SPARTACUS — Stochastic particle solver for multiscale nonequilibrium gas flows based on SPARTA.
- MSPD — the implementation of the Multiscale Stochastic Particle (MSP) method for simulating diatomic gas flows.
- DHC-GEP — Dimensional homogeneity constrained gene expression programming for discovering governing equations.
- SITE — General data-driven framework for identifying tensor equations.
- PINN-DVM — Combination of Physics-Informed Neural Networks (PINNs) and the Discrete Velocity Method (DVM) for the linearized Boltzmann–BGK equation.