Abstract:
|
Reliable design of turbomachinery is a complex task. In order to gain overall
efficiency in the machine performance, designers may have to reduce the gap between the impeller and the diffuser, forcing them to be as closely spaced as possible. In these situations, there may be a strong interaction between them that influences both the aerodynamics and the structural performance of blades and vanes. This phenomenon is called rotor-stator interaction
(RSI), and it has a strong influence on the machine behavior. These interactions can have a significant impact on the vibrational and acoustical characteristics of the machine. Sometimes, this interaction has led to blade or vane failure. Unsteadiness and turbulence play a fundamental role in RSI, and the use of computational fluid dynamics (CFD) is becoming a usual requirement in turbomachinery design due to the difficulties and elevated cost of the experimentation required to identify RSI
phenomena. Nowadays, a CFD analysis based on Reynolds–averaged Navier–Stokes equations (RANS) and a coupled eddy viscosity turbulence model (EVM) is commonly applied in turbomachinery design. Therefore, the choice of an appropriate turbulence model
and the boundary layer treatment is far from trivial, and a suitable turbulence modeling plays an important role for successful CFD results.In this work, an entire stage of a diffuser pump was modeled by means of a commercial CFD code in order to study the pressure fluctuations due to the interaction between the impeller and the diffuser of the pump. The obtained numerical results were compared against the experimental results of Tsukamoto et al.,. Full RANS equations coupled with several EVM were solved for a diffuser pump stage in order to establish the most accurate modeling strategy for a diffuser pump. Boundary layer sensitivity tests were performed, and numerical
discretization influence on results was also tested and established. Frequencies of the pressure fluctuations in the diffuser passage are also obtained with several EVM and compared against experimental results. |