论文部分内容阅读
With the huge growth in high-performance computational resources, in order to remarkably prompt the contribution of turbulent flow simulation in real-life computational fluid dynamics (CFD) applications, it is important to quantify epistemic uncertainty in implementing simulations in terms of using Reynolds-Averaged Numerical Simulations (RANS), Large-Eddy Simulation (LES), hybrid RANS/LES and Direct Numerical Simulations (DNS) in many complex fluid flows.In CFD simulations, it is common that a large number of choices in setting up computational configurations should be made to mimic the real-life flow physics, which include the physical models and related arbitrary constants, boundary conditions, initial conditions as well as tunable parameters in employed numerical methods.In most cases, it is impossible to exactly known all physics and geometrical and sometime, these are even not known at all.These issues inevitably result in existence of uncertainties in complex turbulent flow simulations.To account for this lack of information in simulations, uncertainties in related models must be handled.This is the purpose of Uncertainty Quantification (UQ) theory, which allows for a quantitative description of the system response spanned by possible variation of uncertain parameters.The classical deterministic solutions are replaced by stochastic ones where a continuous description of the space of possible solutions spanned by uncertain parameters is recovered.Meanwhile, by virtue of the process of UQ, some possible new insights about the physics of the simulated system can also be achieved.The epistemic uncertainty can be reduced by incorporating existing data from experiments into the simulation, thanks to Data Assimilation (DA) techniques.In this talk, we provide a survey of recent progress made in the field of uncertainty and error quantification and propagation in turbulent flow simulations.All issues addressed are relevant to four fundamental methodologies: RANS, LES, hybrid RANS/LES and DNS.Finally, the talk concludes with challenges and open issues of UQ in turbulent flow simulations.