Abstract

Turbines and pumps dealing with incompressible flow are examples of hydraulic turbomachines. In most cases the flow is highly turbulent and time-dependent, caused by the rotation of the impeller in a stationary casing. The geometry, with doubly curved surfaces, adds even more to the complexity. It all leads to a flow which is difficult to model. Yet, to optimize turbomachines it is necessary to analyze the flow in detail. Flow simulations using Computational Fluid Dynamics (CFD) can be a very helpful tool. The software solves the discretized partial differential equations for mass and momentum conservation on a grid that covers the flow domain. Two basic discretization schemes can be distinguished: collocated and staggered. When a collocated scheme is used, the solution suffers from odd-even decoupling. In practice this is suppressed with artificial measures which either decrease the accuracy of the simulation or increase the calculation time for an unsteady incompressible flow. Using a staggered scheme, accurate discretization is more difficult, but odd-even decoupling is avoided. In this thesis a CFD code is developed which is based on a staggered, blockstructured grid scheme. It is suited for the calculation of time-dependent fluid motion in turbomachines. The CFD code, named DEFT, is originally developed by the group ofWesseling at Delft University of Technology. The first extension in the current work was an interpolation procedure implemented to handle non-matching grids for more flexibility in grid generation. Furthermore, a sliding interface to connect the rotating grid in the impeller and the stationary grid was developed. Coriolis and centrifugal forces for calculations in the rotating frame of reference, were mplemented in two ways: using a conservative formulation and using source terms. An adaptation of the pressure equation proved necessary to reduce calculation time for computations involving a sliding interface. Although the conceptual ideas behind these extensions are applicable in 3D, they have been implemented in 2D and verified with the simulation of a number of relatively simple flows. DeFT was validated with the simulation of the flow through a cascade of blades which is a model of an axial-flow pump. The blade surface pressure and the total force on the blade are calculated. There is good agreement between values calculated with DeFT, Fluent, values from experiments, and other CFD calculations obtained from literature. The flow through a centrifugal pump with a vaned diffusor is simulated using the staggered discretization in DeFT and the collocated discretization in Fluent. The calculated time-averaged pressure and velocity along the pitch of a rotor channel show good correspondence. The agreement with results from experiments and other CFD calculations obtained from literature is more qualitative. The calculation time needed by DeFT and Fluent is approximately equal, despite the use of a large number of blocks in DeFT and its lack of a convergence enhancing multi-grid method which is used by Fluent.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call