Pump performance prediction using potential flow method

1. Introduction
   
2. PumpFlow
   
3. Examples
   
4. Publications
   

Introduction

The design of hydraulic turbomachines, such as pumps and water turbines, has reached the stage were improvements can only be achieved through a detailed understanding of the internal flow. The prediction of the flow in such equipment is very complicated due to the three-dimensionality of the flow and the highly curved passages in rotating impellers. Furthermore, turbomachines show unsteady flow behaviour, especially under off-design conditions, as a result of the interaction between impeller and volute or stator. Considering these complexities, computer simulations of the flow are becoming increasingly important. Over the past few years, there is a tendency towards the development of numerical methods based on the Navier-Stokes equations, in order to account for viscous effects like wakes, boundary layers and separation bubbles. An open problem in such computations is the choice of an appropriate turbulence model, since the standard turbulence models appear not to be adequate in rotating systems. Furthermore, the computer time needed to compute the flow through a single impeller channel is enormeous, even when using supercomputers. As part of a design tool, these methods are of limited suitability.

Assumptions

Fortunately, the physical model can be simplified considerably without losing its overall validity. As a first assumption the fluid can be considered incompressible. Secondly, the bulk of the fluid can be regarded inviscid, as viscous forces are negligible when compared to inertia forces. Viscous effects are restricted to boundary layers and wakes behind the blades. The third assumption is that the flow enters the impeller free of vorticity. For an inviscid bulk flow this means that the flow remains irrotational. Combining these assumptions leads to the the incompressible potential flow. Many investigators have presented numerical methods for computing the potential flow inside a rotor channel in two or three dimensions, and in complete rotor-stator configurations in two dimensions. In order to impose the Kutta condition of smooth flow at the blades' trailing edges they either superimpose a number of subpotentials (reflecting unit flowrate, unit rotation and line vortices of unit strength, shed from points at the trailing edges) or determine the (varying) circulations iteratively.

Multi-block method

In order to reduce computing time and memory requirements as much as possible, an entirely new potential flow solver has been developed. It is capable of computing the unsteady three-dimensional potential flow inside a rotor-stator configuration, taking into account the varying circulation along the blades' span. The method is based on the Finite Element Method. Two techniques are responsible for its high efficiency: the method of substructuring (multi-block method) and the implicit imposition of Kutta conditions at the blades' trailing edges.

Without going to much into detail, the method is as follows. The flow region of interest is divided into subdomains or blocks, all having a topologically cubic shape (see the above figure). Blocks in the rotor of the pump are separated from blocks in the stator by a cylindrical or comical surface. Thus the rotation of the impeller with respect to the stator can be simulated by sliding the rotorblocks along this surface. A structured grid is generated for each of the blocks (figure below). The solution method consists of two steps. In the first one, internal degrees of freedom will be eliminated from the system of equations, for all blocks separately. The remaining degrees of freedom at internal block boundaries are then coupled in the second step. Kutta equations which prescribe flow conditions at trailing edges are added. After solving the resulting system of equations, the previously eliminated degrees of freedom are finally obtained.

PumpFlow

automated gridgenerator based on a parametric impeller geometry description

automated gridgenerator for single volutes with a trapezoidal cross-section

multi-block potential flow solver

postprocessor

visualization system

graphic user interface runs on Windows 3.1/3.11, Windows95, en Linux platforms

flow solver can be run separately on a larger/faster computer (mainframe, workstation, supercomputer, ...)

saving of time through faster input and reduction of errors

interactive postprocessing

possibility to compute user defined quantities easily

integrated visualization system

possiblity of importing meshes from other gidgeneration programs

possibility of generating input files for third party visualization software, like AVS and POVRAY

Examples

PUBLICATIONS

RELATED PUBLICATIONS

To view pdf-files, download the latest version of Acrobat Reader from Adobe

For comments and questions regarding this page email to b.p.m.vanesch@wtb.tue.nl

Last update: June 22, 1999