See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258683038

Numerical simulation of centrifugal casting of pipes

Article  in  IOP Conference Series Materials Science and Engineering · July 2012

DOI: 10.1088/1757-899X/33/1/012031

CITATIONS READS

8 504

1 author:

Erhard Kaschnitz
Österreichisches Gießerei-Institut
56 PUBLICATIONS   953 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Erhard Kaschnitz on 20 May 2016.

The user has requested enhancement of the downloaded file.

Home Search Collections Journals About Contact us My IOPscience

Numerical simulation of centrifugal casting of pipes

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2012 IOP Conf. Ser.: Mater. Sci. Eng. 33 012031

(http://iopscience.iop.org/1757-899X/33/1/012031)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 213.159.32.110
This content was downloaded on 11/03/2016 at 19:42

Please note that terms and conditions apply.

MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

Numerical simulation of centrifugal casting of pipes

E Kaschnitz1
Österreichisches Gießerei-Institut, Parkstraße 21, 8700 Leoben, Austria

E-mail: [email protected]

Abstract. A numerical simulation model for the horizontal centrifugal pipe casting process
was developed with the commercial simulation package Flow3D. It considers - additionally to
mass, energy and momentum conservation equations and free surface tracking - the fast radial
and slower horizontal movement of the mold. The iron inflow is not steady state but time
dependent. Of special importance is the friction between the liquid and the mold in connection
with the viscosity and turbulence of the iron. Experiments with the mold at controlled
revolution speeds were carried out using a high-speed camera. From these experiments friction
coefficients for the description of the interaction between mold and melt were obtained. With
the simulation model, the influence of typical process parameters (e.g. melts inflow, mold
movement, melt temperature, cooling media) on the wall thickness of the pipes can be studied.
The comparison to results of pipes from production shows a good agreement between
simulation and reality.

1. Introduction
Ductile iron water pipes and piles for deep foundation engineering are produced using the “de
Lavaud” horizontal centrifugal pipe casting technique. In this process, the liquid iron is filled through
an open channel into a fast rotating hot-work tool-steel mold that is slightly tilted. The tool-steel is
water-cooled at the outside. In order to distribute the metal, the spinning mold including its cooling
system is shifted axially in a controlled movement.
The numerical simulation of this process is challenging, no description of its numerical simulation
was found in literature. Some papers describe the less complex vertical centrifugal casting process
[1-6]. Besides taking into account the mass, energy and momentum conservation equations and the
free surface tracking, the following phenomena have to be considered for the horizontal centrifugal
casting process of pipes: (i) relatively fast radial movement of the mold; (ii) relatively slow but time
dependent movement of the mold in axial direction; (iii) time dependent inflow of the iron along the
channel; (iv) friction between the liquid and the mold; (v) turbulence and viscosity of the iron. The
situation is hindered by an unfavorable relation between the relatively thin wall thickness of the tube
(in the millimeter range) and its overall length (several meters).
After overcoming instability difficulties, it was possible to develop a valid model with Flow3D. In
this work, the basic principles of the solution method are discussed; examples of numerical
simulations and practical experiments are presented.

1
To whom any correspondence should be addressed.

Published under licence by IOP Publishing Ltd 1

MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

2. Simulation model
The geometry of the model consists of a mold from tool-steel and a closing core from cold-box
bonded silica sand, both parts are spinning and moving (figure 1). An open feeding channel is
extended into the mold, this channel is fixed. The outer side of the mold is water-cooled (significant
heat flow), the inner side contains air (almost no heat loss).

Figure 1. Vertical cut through the geometry of the model (1,

mold; 2, sand core; 3, open feeding channel). The arrows show
the direction of the movement of mold and core.

The mold and the core (but not the open channel) are cylindrical. As the melt flow follows the
inner surface of the mold most of the time, a cylindrical coordinate system was chosen. In comparison
to a regular Cartesian coordinate system, most of the unfavorable flow between neighboring cells
under an angle of 45 degree is avoided.
For faster and better computations, the (cylindrical) mesh is aligned to the mold that is slightly
tilted. This means the gravity vector gets the same tilt in the axial direction and splits in two
components. Using the cylindrical mesh, the so-called “Non-inertial-reference-frame”-model of
Flow3D has to be introduced to be able to define the gravity vector appropriately [7].
The relatively slow but time dependent movement of the mold is described by the “General-
moving-obstacles”-model of Flow3D. This model shifts the assigned obstacles (mold and core)
according to the pre-defined motion. Additional source terms according to the loss of mass,
momentum and energy are added to the conservation equations at mesh cells in the vicinity of the
shifted obstacles to compensate for these cells now covered by the obstacle [7].
This approach is not suitable for the relatively fast spinning of the mold and core, because the time
steps become increasingly smaller with increasing spin velocity. Meaningful simulations would take
months or years. For the rotational movement of mold and core, the so-called “Ospin”-model of the
software package Flow3D was used. This model does not spin the obstacle itself, but transfers an
appropriate amount of momentum as shear force from the obstacle in contact with the melt (inner
surface of the mold and outer surface of the core) to the melt itself. The simulation model is
comparable to a merry-go-round, where the motionless person outside pushes the child in the seat for
rotational movement.
The free surface and solidification behavior of the fluid was modeled with Flow3D in a standard
way. But Flow3D does no momentum and mass conservation calculations for already solidified cells.
For this reason, as soon as the melt in a cell is solidified, the rigid cell is moved artificially by the

2
MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

predefined rotational and horizontal movement. The rotation speed leads only to a limited centrifugal
pressure in the melt (the wall-thickness is rather small): the solidification behavior is not assumed to
be influenced by the rotation.
The simulated pipes have a wall thickness of approximately five millimeter and are several meter
long. This means that even big meshes with long computation times (up to three weeks) are relatively
coarse. For that reason, special care has to be taken to choose appropriate values for viscosity and the
turbulence model as well as the amount of shear stress that is transferred from mold to melt. A
Renormalization-Group method is used for the transport equations for turbulent energy and its
dissipation (k-epsilon model). A surface roughness factor provides additional shear stress between
wall and the first cell layer of metal. These parameters are slightly dependent on the relatively rough
mesh. This dependency should vanish for better meshes, but these are not to compute in reasonable
time-spans nowadays. In order to approximate the fluid flow with correct parameters, practical
experiments were carried out.

3. Experiments
The experimental setup consists of an electrical engine driving a horizontal cylindrical mold at
precisely controlled revolution speed [8]. The control provides either constant speed or an angular
velocity ramp at constant acceleration. The mold used for the experiments was a 300 mm piece of an
original 5 meter hot-work tool-steel mold with its original inner surface. The sample mold has two lids
at the front and at the backside; the front side lid is transparent to be able to watch the fluid inside
(figure 2).

Figure 2. Experimental setup for the spinning experiments (1, mold; 2, engine; 3 gear; 4, high-speed
camera).

Two different fluids were used for the experiments: colored water at room temperature and Rose's
alloy (Bi, 50 wt-%; Pb, 25 wt-%; Sn, 25 wt-%) at approximately 100°C. The behavior of the fluid at
different revolution speeds and different acceleration ramps was filmed from the front by a high-speed
digital camera at 600 frames per second.
Figure 3 shows the shape of the free surface of the metal at a certain revolution speed. As the
contrast of high-speed movies is poor in principle (and even worse in a printed paper), the interface
position in the figure is marked by red arrows. The metal undergoes three phases during acceleration:
(i) it forms a wedge shaped surface along the cylinder; (ii) the free surface decays and a curtain forms;
(iii) when the angular velocity is high enough, the looping condition is satisfied and suddenly the
curtain vanishes and the metal turns stable along the inner surface of the mold. This behavior depends
mainly on viscosity of the melt and the surface condition of the mold. Figure 4 shows two simulations
with different surface roughness, in one case the melt is unstable and forms a curtain, the other case
shows a stable spinning of the melt.

3
MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

In a second experimental setup, the engine and gear system was turned by an angle of 90 degree to
use a part of the mold as a turntable. With that setup, liquid cast iron could be used instead of Rose's
alloy. Using the gained knowledge from the experiments, meaningful input parameters were found for
the simulation of the specific industrial pipes.

Figure 3. Shape of the free surface of the Rose’s alloy at

a selected revolution speed.

Figure 4. The simulation with low surface roughness (left) shows an unstable melt
curtain, the simulation with an appropriate surface roughness (right) shows a stable
melt following the looping condition.

4
MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

4. Results
Figure 5 shows an example of the temperature distribution during filling of a pipe in a front view with
mold and core faded out.
The wall-thickness of the produced pipes should be as uniform as possible. Therefore, the most
interesting simulation result is the predicted wall-thickness of a pipe along its overall length. From
each individual simulation with specific production parameters, the position of each fluid cell, its size
and fill fraction was read out from Flow3D in order to compute the wall-thickness along the length.
Depending on production parameters, the shape of the pipe can be quite different. As an example,
figure 6 shows the wall-thickness of several pipes as a function of length.

Figure 5. Simulated temperature distribution during filling of a pipe in a front

view. The hot inflowing melt (in red color) is distributed along the inner tube-
shaped steel mold due its rotation.

5. Conclusions
Following conclusions can be drawn:
The De-Lavaud process can be simulated with Flow3D, taking advantage of the manifold
modeling capabilities of this software
Long calculation times (15 to 20 days) per pipe are common
In comparison to reality, the influence of the main process parameters can be reproduced in
the simulation
Computed wall thickness matches within the relatively coarse grid size with measurements of
standard pipes
In simulation, each single process parameter can be varied and its influence on wall thickness
can be studied
Combinations of process parameters and hard (or not) to change parameter sets were
simulated
The impact of important parameters was identified.

5
 MCWASP XIII IOP Publishing
IOP Conf. Series: Materials Science and Engineering 33 (2012) 012031 doi:10.1088/1757-899X/33/1/012031

12

10
Wall thickness, mm

0
0 1 2 3 4 5

Length of pipe, m
Figure 6. Examples for several simulated wall-thicknesses along the length of
pipes using different process parameter sets.

References
[1] Wu S P, Liu D R, Guo J J, Li C Y, Su Y Q, Fu H Z 2006 Mat. Sci. Eng. A-Struct 426 240-9
[2] Fu P X, Kang X H, Ma Y C, Liu K, Li D Z 2008 Intermetallics 16 130-8
[3] Gao J W, Wang C Y 2000 Mat. Sci. Eng. A-Struct 292 207-15
[4] Wu S P, Li C Y, Guo J J, Su Y Q, Lei X Q, Fu H Z T 2006 Nonferr. Metal. Soc. 16 1035-40
[5] Shevchenko D M, McBride D, Humphreys N J, Croft T N, Whitey P, Green N R, Cross M 2009
Centrifugal casting of complex geometries: computational modelling and validation
experiments Modeling of Casting, Welding and Advanced Solidification Processes - XII eds S L
Cockroft and D M Maijer (Warrendale, PA: The Minerals, Metals & Materials Society (TMS))
pp 77-84
[6] McBride D, Humphreys N J, Croft T N, Whitey P, Green N R, Cross M 2009 Complex free
surface flows for mould filling using centrifugal casting Modeling of Casting, Welding and
Advanced Solidification Processes - XII eds S L Cockroft and D M Maijer (Warrendale, PA:
The Minerals, Metals & Materials Society (TMS)) pp 459-66
[7] 2008 Flow3D Version 9.3 – User Manual (Santa Fe, NM: Flow Science Inc.)
[8] Lackner R 2011 Diploma thesis (Leoben, Austria: Montanuniversität Leoben)

View publication stats