Requirements for Large Eddy Simulation Computations of Variable-Speed Power Turbine Flows

Variable-speed power turbines (VSPTs) operate at low Reynolds numbers and with a wide range of incidence angles. Transition, separation, and the relevant physics leading to them are important to VSPT flow. Higher fidelity tools such as large eddy simulation (LES) may be needed to resolve the flow features necessary for accurate predictive capability and design of such turbines. A survey conducted for this report explores the requirements for such computations. The survey is limited to the simulation of two-dimensional flow cases and endwalls are not included. It suggests that a grid resolution necessary for this type of simulation to accurately represent the physics may be of the order of Δx+=45, Δx+ =2 and Δz+=17. Various subgrid-scale (SGS) models have been used and except for the Smagorinsky model, all seem to perform well and in some instances the simulations worked well without SGS modeling. A method of specifying the inlet conditions such as synthetic eddy modeling (SEM) is necessary to correctly represent the inlet conditions. Introduction and Motivation Variable-speed power turbines (VSPTs) for rotorcraft applications operate at low Reynolds numbers and a wide range in incidence resulting from rotational speed variation (Welch, 2010). At all speeds, the blades operate in conditions like those in highly-loaded low pressure turbine (LPT) blades that have been the subject of research in recent years. The flow physics that are important to LPT operation resulting from the operating conditions, that is, separation, transition, and the resulting losses, are fully expected to apply to VSPTs. Blade designs are such that the on-design flow on the suction side experiences a vigorous acceleration and then goes through a region of rapid deceleration. The acceleration parameter K defined as (ν/U 2) dU/dx typically ranges ±5.0×106 for a Reynolds number of 105 which is large compared to, for example, high pressure turbine since K is scaled by the Reynolds number (Mayle, 1991). Because of the presence of free-stream turbulence (FST) in the turbines and passing wakes from upstream blade rows and their associated turbulence, the process of transition is not natural. Background turbulence, even at low levels, discourages natural transition. The boundary layer on the suction side of the blades is initially laminar due to acceleration, but it can transition given high enough turbulence intensity. If the boundary layer does not transition in the initial accelerating, or mildly decelerating region on the suction side, the layer is prone to separation in the strongly decelerating flow regime and transition will take place upon reattachment. Efficiency lapses observed in LPTs at cruise conditions (low Reynolds number condition) is caused by separated flow in the LPT blades. Reattachment before the trailing edge would ensure lower losses while an open separation bubble leads to larger losses (Medic and Sharma, 2012). Reattachment length itself is a strong function of FST (Mayle, 1991). At sufficiently high operating rotational speeds with the associated negative incidence, the VSPT flow separates from the pressure surface and a separation and turbulent reattachment takes place on the pressure surface. It is obvious that development of the boundary layer with the contributing factors such as the effects of turbulence and