3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in

Hydraulic Machinery and Systems, October 14-16, 2009, Brno, Czech Republic

The ERCOFTAC centrifugal pump OpenFOAM case-study

Maryse PAGE
Hydro-Qubec, Institut de recherche, Varennes, Canada

Martin BEAUDOIN
Hydro-Qubec, Institut de recherche, Varennes, Canada

Hkan NILSSON
Chalmers University of Technology, Gteborg, Sweden

ABSTRACT
This work investigates the rotor-stator interaction features of OpenFOAM-1.5-dev, such as frozen
rotor and sliding grid. The case studied is the ERCOFTAC Test Case U3: Centrifugal Pump with a
Vaned Diffuser, a testcase from the ERCOFTAC Turbomachinery Special Interest Group. The case
was presented by Combs at the ERCOFTAC Seminar and Workshop on Turbomachinery Flow
Prediction VII, in Aussois, 1999. It is a valid test case for evaluation of rotor-stator interaction
features, as detailed experimental data is available.
The investigation shows that OpenFOAM gives results that are comparable to the experimental data,
in particular for the sliding grid case. The results are less accurate in the frozen rotor simulation due to
the improper treatment of the impeller wakes that is part of the frozen rotor formulation.
The ERCOFTAC centrifugal pump OpenFOAM case-study was developed as a contribution to the
OpenFOAM Turbomachinery Working Group, and was presented and discussed at the Fourth
OpenFOAM Workshop in Montral, 2009. The complete set-up of the case-study is available from the
OpenFOAM-extend project at SourceForge, and instructions and comments are available from the
OpenFOAM Wiki.

KEYWORDS
CFD, OpenFOAM, Turbomachinery, Frozen rotor, Sliding grid, GGI, ERCOFTAC centrifugal pump

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

1 INTRODUCTION
OpenFOAM is an Open Source library written in C++ [1]. It is a well-structured code, mostly
used to implement CFD solvers, although it is also used in other applications. OpenFOAM is
based on the finite volume method, but there are also implementations of the finite area and
finite element methods. The code accepts fully unstructured meshes and polyhedral cells.
Many advanced features can be found in OpenFOAM, such as moving meshes and conjugate
heat transfer. With regards to basic features, such as turbulence models and discretization
schemes, OpenFOAM is a serious and high quality CFD tool that is constantly evolving.
The community-driven OpenFOAM Turbomachinery Working Group [2] develops and
validates OpenFOAM for turbomachinery applications. The ERCOFTAC centrifugal pump
was chosen as a validation test case for this investigation of the rotor-stator interaction
features in OpenFOAM. The results of the initial studies of this case were presented at the
Fourth OpenFOAM Workshop in June 2009, in Montral, Qubec. All the files are available
at the OpenFOAM-extend project at SourceForge, for anyone who would like to learn
OpenFOAM, or become familiar with the turbomachinery features in OpenFOAM. Some of
the important features, such as the GGI (General Grid Interface), multiple frames of reference,
and moving meshes are described in the present paper. The GGI forms the base for many
useful features for turbomachinery.
2 METHOD
In this work, the flow through a centrifugal pump is investigated. The incompressible
Reynolds-Averaged Navier-Stokes equations are solved, using the finite volume method and a
standard k- turbulence model closure with wall-functions. The convection discretization uses
a second-order linearUpwind scheme. Two different rotor-stator approaches are used:
The frozen rotor approach is a steady-state formulation where the rotor and stator are
fixed with respect to each other, and different reference frames are used in the rotating
and stationary parts. This is also referred to as Multiple Reference Frames (MRF), and
it allows taking into account the effect of the rotation of the impeller, although no
transient rotor-stator interaction is included. It is nevertheless a fast preliminary
method, for use as initial conditions for sliding grid simulations.
The sliding grid approach is a transient method where the rotor mesh actually rotates
with respect to the stator mesh. The interaction between the rotor and stator are thus
fully resolved. This requires a sliding grid interface between the rotor and stator
domains. In OpenFOAM there are two alternatives for sliding interfaces, GGI and
topology changes.
Details on the two approaches can be found in section 4.
3 THE ERCOFTAC CENTRIFUGAL PUMP CASE-STUDY
The ERCOFTAC centrifugal pump is a simplified model of a centrifugal turbomachine. The
original test case was presented by Combs [3] at a Turbomachinery Flow Prediction
ERCOFTAC Workshop in 1999.

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

3.1 Geometry
The simplified model of the centrifugal pump has 7 impeller blades, 12 diffuser vanes and 6%
vaneless radial gap as shown in Fig. 1. The geometric data and operating conditions are
shown in Tab. 1. The test rig was built by M.Ubaldi et al.[4]. The purpose was to study the
flow unsteadiness generated by rotor-stator interaction in a turbomachine.

Fig. 1: Impeller and vaned diffuser geometry (Image taken from Ubaldi et al. [4]).
Impeller

Diffuser

Operating conditions

Leading edge
diameter

D1 = 240mm

Leading edge
diameter

D3 = 444mm

Rotational
speed

n = 2000rpm

Trailing edge
diameter

D2 = 420mm

Trailing edge
diameter

D4 = 664mm

Impeller tip
speed

U2 = 43.98m/s

Number of blades

zi = 7

Number of vanes

zd = 12

Flow rate
coefficient

Blade span

z = 40.4mm

Outlet diameter

D5 = 750mm

Total pressure
rise coefficient
Reynolds
number
Air density

=
=

4Q
= 0.048
U 2 D22

2( p out pin )
= 0.65
8 U 22

Re =

U 2l
= 6.5 10 5

= 1.2kg/m3

Tab. 1: Geometric data and operating conditions [4].

3.2 Measurements

The experimental data was provided by Ubaldi et al. [4]. The model operates in an open
circuit with air directly discharged into the atmosphere from the radial diffuser. The pump
operates at the nominal operating condition, at a constant rotational speed of 2000 rpm.
(Reynolds number: 6.5x105, incompressible flow regime). Phase locked ensemble averaged
velocity components have been measured with hot wire probes at the impeller outlet. The data
includes the distribution of the ensemble averaged static pressure at the impeller front end,
taken by means of miniature fast response transducers mounted at the stationary casing of the
impeller. LDV measurements were also performed in the impeller and in the diffuser.
The measurements made by Ubaldi were made using a hot-wire probe at a radial distance 4
mm from the blade trailing edge (Dm/D2=1.02). The results from the simulations were
extracted in a similar way, as shown in Fig. 2.

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

The simulated results are

Fig. 2: Position of the hot-wire probe, and of the sampling of the simulated data.

3.3 Case-study set-up and cases

The ERCOFTAC centrifugal pump was used as a case study for the Fourth OpenFOAM
Workshop, held in Montral, Qubec, in June 1-4 2009. It can be found in the turbomachinery
Special
Interest
Group
web
page
at
the
OpenFOAM
Wiki
(http://openfoamwiki.net/index.php/Sig_Turbomachinery), and in the OpenFOAM-extend
SourceForge project (http://sourceforge.net/projects/openfoam-extend/). It includes blockstructured meshes, scripts that automatically set up the cases, and automatic post-processing
of the results and comparison with the measurements. There are also written descriptions of
the OpenFOAM features that are used for the case-study.
In the present work, a 2D representation of the geometry is used. Previous simulations have
been done by Bert, Combs and Kueny [5], who showed that relevant information could be
recovered from 2D simulations although the real flow has 3D features. The 2D mesh was
made using ICEM-HEXA, and the rotor and the impeller were meshed separately. The mesh
is block-structured, and consists of about 94 000 cells, with an average Y+ value of 35.
The boundary conditions are shown in Tab. 2.
Calculated data for the 2D cases

Boundary conditions

Vradial = U0
Inlet Diameter

Z thickness

Z = 1mm

(OpenFOAM requires one cell

Flow rate
Inlet radial speed

T
= 10 (viscosity ratio )

D0=200 mm

At the inlet

U 2 D22
Q=
= 0.292m 3 / s
4
U0 =

Q
Q
=
= 11.4m / s
A0 2r0 z

3
k = U 02 I 2 = 0.48735m 2 / s 2
2
(I=5%)
=

At the outlet

C k 2
T

C k 2

( T / )

C k 2

( T / )

Average static pressure 0

Tab. 2: Computational parameters for the 2D cases.

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

The ERCOFTAC centrifugal pump (ECP) cases that are currently available are listed here:
Steady state frozen rotor cases, using MRF (Multiple Reference Frames)

ECPGgi2D: The frozen rotor approach, and the GGI between the impeller and the
diffuser.

ECPStitchMesh2D: The frozen rotor approach, where the rotor and stator meshes
have been stitched together at the interface, forming a single mesh with hanging
nodes at the interface.

Unsteady sliding grid cases

ECPMixerGgiFvMesh2D: The sliding grid approach, where the GGI is applied

ECPMixerFvMesh2D: The sliding grid approach, where the rotor and stator
meshes are stitched together at the interface at each time step, forming a single
mesh with hanging nodes at the interface at each time step.

3.4 Results
3.4.1 Steady-state simulation frozen rotor

The simulation was stopped after 5000 iterations, since all the residuals were below 10-5. The
frozen rotor approach gives something that resembles a snapshot of the real flow in the pump,
but the advection of the impeller wakes in the diffuser region will by definition not be
physical (see Fig. 3).

Fig. 3: Velocity magnitude and static pressure for the pump, for the frozen rotor simulation.

The computed velocities show some similarity with the experimental data, as shown in Fig. 4,
however, the results do not agree perfectly with the experimental ones. Most of these
differences are likely due to the frozen rotor formulation rather than the OpenFOAM
implementation. A comparison with preliminary results from a commercial CFD code [6]
shows that OpenFOAM gives similar results as commercial CFD codes for frozen rotor
simulations.

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

The GGI and the stitched cases give exactly the same results as can be seen in Fig. 4, and it
can thus be concluded that the interface coupling works as it should.
-0.2
-0.3

experimental results
stitch simulation
GGI interface simulation

0.25
0.2

-0.4
Radial velocity

Tangential velocity

0.3

experimental results
stitch simulation
GGI interface simulation

-0.5
-0.6

0.15
0.1

-0.7

0.05

-0.8

0
-0.05

-0.9
0

0.5

Abscissa

1.5

0.5

Abscissa

1.5

Fig. 4: Distributions of the tangential and radial velocities, for the two frozen rotor simulations
compared to the experimental data. The two simulation results are identical.

3.4.2 Unsteady simulation, using the sliding grid approach

Both computations (ECPMixerGgiFvMesh2D and ECPMixerFvMesh2D) give the same

Fig. 5: Velocity magnitude and static pressure for the ERCOFTAC centrifugal pump, for the unsteady simulation

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

-0.2
-0.3

experimental results
computational results

0.25
0.2

-0.4
Radial velocity

Tangential velocity

0.3

experimental results
computational results

-0.5
-0.6

0.15
0.1
0.05

-0.7

0
-0.8

-0.05
0

0.5

Abscissa

1.5

0.5

Abscissa

1.5

Fig. 6: Distributions of tangential and relative velocity for the unsteady simulation, compared to the
experimental data.

4 IMPLEMENTATION DETAILS
4.1 Steady-state solver for Multiple Reference Frames

The solver used for the frozen rotor simulations, based on the Multiple Reference Frame
(MRF) approach is named MRFSimpleFoam. It is a steady state solver for incompressible
turbulent flow, using the SIMPLE algorithm for pressure-velocity coupling. The MRF
approach implies that there is no relative mesh motion of the rotating and stationary parts. In a
rotating reference frame, where the relative velocity is computed, the momentum equations
must be modified with Coriolis and centrifugal terms. In the MRF approach, the momentum
equations uses a mix of inertial and relative velocities, and only one extra term appears in the
equations. Tab. 3 summarizes the formulation of the Reynolds Averaged Navier-Stokes
equations in the inertial and the rotating frames, and the third alternative is what is
implemented in MRFSimpleFoam.
Frame

Convected velocity

Inertial

Absolute velocity

Rotating

Relative velocity

Rotating

Absolute velocity

Steady incompressible Navier-Stokes equations

Tab. 3: Summary of the Navier-Stokes equations for steady flow in multiple reference frames.

4.1.1 Frozen rotor

In the frozen rotor formulation, the rotating and stationary parts are considered to be at a fixed
position relative to each other. The coupling between the rotor and stator domains are still
resolved 360, but fixed in time. Since the rotor and the stator parts have been meshed
separately, a connection must be made between these meshes. Two methods can be used for
that in OpenFOAM. The first one is based on the topological changes technology (see section
4.3). The second method is the GGI (see section 4.4).

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

4.1.2 Mixing plane

Another solution to model the interface between the rotating and stationary parts is the mixing
plane. At the interface the flow properties are circumferentially averaged. This will of course
remove all transient rotor-stator interactions, but it still gives fairly representative results. A
mixing-plane simulation only requires one rotor blade and one stator blade per stage, which
significantly accelerates the solution procedure.
The mixing plane is under implementation in OpenFOAM at the moment, but there should be
an implementation available in the coming year.
4.2 Unsteady solver for sliding grid

The solver used for the sliding grid simulations in this work is named turbDyMFoam. This is
a solver for incompressible RANS simulations using the PISO algorithm for pressure-velocity
coupling. The solver also uses libraries for mesh motion and deformation of polyhedral
meshes [7], of which one of them handles partly rotating meshes. In the simulations the
physical motion of the mesh is directly addressed. The coupling between the rotating and nonrotating parts of the mesh can be accomplished by either topological changes (see section 4.3)
or GGI (see section 4.4).
4.3 Topological changes

Two parts of a mesh with coinciding faces can be attached to each other by connecting the
faces of both parts. If the resulting mesh would not be conformal, OpenFOAM allows
polyhedral cells, and may thus cut the faces to make the mesh conformal with hanging nodes.
The sliding interface with topological changes in OpenFOAM uses this functionality at each
time step, and deals with the topological changes associated to it. The topological modifier
attach-detach is then used, and the rotating part of the mesh is detached from the static part.
The rotation then occurs, and the mesh is attached again, as explained in Fig. 7. The black
points show the vertices of a 2D cell, and it is shown that a non-conformal connection gives
control volumes with hanging nodes.

Fig. 7: Operation of a sliding interface, pictures from [7].

This operation can be quite time-consuming, as it needs to re-organize the topology and the
internal numbering of faces and cells at each time step. However, the method is fully
conservative.
4.4 General Grid Interface (GGI)

The GGI can be used to speed up the operation of the coupling between the rotor and stator
domains. With the GGI, the modus operandi is quite similar as for the topological change
method. The neighbours still need to be evaluated, but the attach-detach topological modifier
doesn't need to be called anymore. This is a gain of time at each time step that is very
noticeable on the whole simulation.
Implicit couplings are present in OpenFOAM in order to join multiple mesh regions into a
single contiguous domain. But most of them are built to join conformal mesh regions, where
8

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

the patches nodes on each side of the interface are matching one to one. The GGI, developed
by M. Beaudoin and H. Jasak [8] is a coupling interface used to join multiple non-conformal
regions where the patches nodes on each side of the interface do not match.
A GGI is commonly used in turbomachinery, where the flow is simulated through a
succession of various complex geometries. The requirement to fit all meshes with conformal
matching interface is often very hard to reach. Using the GGI, non-conformal meshes can be
designed separately, and joined together using one of many GGI alternatives.
The basic GGI is using weighted interpolation to evaluate and transmit flow values across a
pair of conformal or non-conformal coupled patches. It is similar to the static sliding
interface, although much simpler in the sense that no re-meshing is required for the
neighbouring cells of the interface. The GGI weighting factors are basically the percentage of
surface intersection between two overlapping faces.
The GGI implementation is using the Sutherland-Hodgman algorithm to compute the master
and shadow face intersection surface area [9]. This algorithm is simple, fast and robust; being
re-entrant, it allows for a very compact source code implementation. The SutherlandHodgman algorithm is also generic enough to handle any convex n-sided polygons.
5 Conclusions

The rotor-stator interaction features of OpenFOAM have been investigated and compared
with experimental data of the ERCOFTAC centrifugal pump with a vaned diffuser. Both
steady state simulations using Multiple Reference Frames and unsteady simulations using a
transient solver with sliding mesh were performed. Good agreements were found with the
experiments, but improvements can be made. The unsteady simulation showed a good
behavior of the wakes being advected through the diffuser. It can be concluded that all the
functionality is available in OpenFOAM for accurate rotor-stator analysis.
In order to improve the numerical results the same simulations will be performed in 3D, and
alternative boundary conditions will be investigated. The coming mixing plane
implementation will also be evaluated in the same case, and a tutorial on how to use the
mixing plane will be released to the OpenFOAM community.
6 ACKNOWLEDGEMENTS

We would like to express our greatest acknowledgements to M. Ubaldi who made the
experimental results of the ERCOFTAC centrifugal pump available, to share with the
OpenFOAM community. We would also like to acknowledge OpenCFD Ltd, as well as
Hrovje Jasak from Wikki Ltd, who distribute OpenFOAM, and helped us a lot to create this
case study. We are very grateful to Hydro-Qubec for its financial support and its
participation in this project. Olivier Petit and Hkan Nilsson are partly financed by SVC (the
Swedish Water Power Center, www.svc.nu). SVC has been established by the Swedish
Energy Agency, ELFORSK and Svenska Kraftnt together with Chalmers University of
Technology, Lule University of Technology, Uppsala University and the Royal Institute of
Technology.

IAHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

7 REFERENCES

[1] Weller H.G, Tabor G, Jasak H, Fureby C., A tensorial approach to computational
continuum mechanics using object-oriented techniques, Computers in Physics, Vol.12, No.6,
1998.
[2] Nilsson H., Page M., Beaudoin M., Gschaider B. and Jasak H., "The OpenFOAM
Turbomachinery Working Group, and Conclusions from the Turbomachinery Session of the
Third OpenFOAM Workshop", 24th IAHR Symposium on Hydraulic Machinery and Systems,
October 27-31, 2008, Foz Do Iguassu, Brazil.
[3] Combs, J.F., "Test Case U3: Centrifugal Pump with a Vaned Diffuser", ERCOFTAC
Seminar and Workshop on Turbomachinery Flow Prediction VII, Aussois, jan 4-7, 1999.
[4] Ubaldi M., Zunino P., Barigozzi G. and Cattanei A., "An Experimental Investigation of
Stator Induced Unsteadiness on Centrifugal Impeller Outflow", Journal of Turbomachinery,
vol.118, 41-54, 1996.
[5] Combs, J.F., Bert, P.F. and Kueny, J.L., "Numerical Investigation of the Rotor-Stator
Interaction in a Centrifugal Pump Using a Finite Element Method", Proceedings of the 1997
ASME Fluids Engineering Division Summer Meeting, FEDSM97-3454, 1997.
[6] Page, M., Throux, E. and Trpanier, J.-Y., "Unsteady rotor-stator analysis of a Francis
turbine", 22nd IAHR Symposium on Hydraulic Machinery and Systems, June 29 July 2,
2004, Stockholm, Sweden.
[7] Jasak, H., Dynamic Mesh Handling in OpenFOAM, 47th AIAA Aerospace Sciences
Meeting Including the New Horizons Forum and Aerospace Exposition, 5-8 January,
Orlando, Florida, 2008 (AIAA 2009-341).
[8] Beaudoin M. and Jasak H., "Development of a Generalized Grid Interface for
Turbomachinery simulations with OpenFOAM", Open Source CFD International Conference
2008.
[9] Sutherland I.E. and Hodgman G.W., "Reentrant polygon clipping, Communication of the
ACM, vol.17, Number 1, 32-42, 1974.

10