Influence of flow patterns on the performance of straight-through labyrinth seals. Numerical study and validation against experimental data

Abstract

When numerically simulating the performance of a complete Francis turbine, it is common practice to omit the seals to save computing power. Volumetric and friction losses are usually taken into account using empirical values. There is abundant literature on the topic, but all analytic formulations rely on widely variable empirical coefficients. To overcome the uncertainty in determining volumetric losses, a detailed analysis of the sealing is necessary. Focusing on straight seals and configurations with cavities, in-house experimental data and numerical simulations have been used to show that few geometric parameters play a fundamental role in defining the flow patterns inside the clearance and even more so inside and between cavities. As is to be expected, the seals effectiveness is heavily influenced by these flow structures. It is also shown that certain methods of dealing with near-wall flow structures are better suited than others for the estimation of the discharge behaviour when compared to experimental results. Even though in the framework of the RANS equations CFD has proven to be a reliable tool for a wide range of geometries, further investigations are necessary on deep cavities. For aspect ratios above 0.75, the flow patterns inside the cavity change drastically, leading to bad convergence rates of the solver and, more important, to a considerable misalignment between numerical and experimental results. Lid-driven cavity flow is known to be a complex problem in 2D domain, and in this case, the flow becomes three-dimensional due to the tangential velocity induced by rotation. However, the reasons behind the problems in CFD analysis are not entirely clear. Given the observed variations of the flow structures inside sealings with different geometric parameters, it becomes clear why a simplified, generic formulation is very difficult to obtain, if not by adopting empirically defined coefficients which are needed to tune the analytical model.