Wall-Resolved Large-Eddy Simulation of the LS89 Cascade Using an Explicit Local Time-Stepping Method

Abstract

Abstract In most turbomachinery CFD simulations, a large heterogenity of mesh refinement exists because specific flow regions are known to require small cell sizes for proper prediction. This leads to large cell volume ratios between the smallest and largest cells. In the context of explicit solvers where the timestep is imposed by stability constraints and smallest cell size, large difference in cell sizes may directly lead to a waste of CPU time since this timestep is much smaller than required in the majority of cells. The introduction of a Local Time-Stepping (LTS) is a potential solution to such an observation. The computational domain is divided in multiple sub-domains and each sub-domain will meet its local stability constraints, relaxing the limit of maximal allowed timesteps. In the following, this technique is applied to a wall-resolved Large-Eddy Simulation (LES) of the LS89 turbine vane cascade dividing the domain in two parts: one part containing the near-wall region of the blade (with prism layers at the blade surface leading to a small timestep) and a second part composed of the remaining volume. To evaluate the strategy, a standard wall-resolved simulation and its LTS counterpart are compared to experimental data. Overall, for the LS89 configuration that is known as challenging, there is an excellent agreement between simulations and experiment. More importantly, a 31% speedup is achieved on this configuration while giving identical results when comparing the single domain and the LTS cases despite the significant simulation cost reduction for the latter.