Xample Tirring ANK: January 2017 13.1 Release 18.0
Xample Tirring ANK: January 2017 13.1 Release 18.0
EXAMPLE 13
STIRRING TANK
DESCRIPTION
In this example, contrary to a previous one (Example 11), we can not simplify the simulation
by assuming a 2-D flow. The blades geometry, as can be seen on Fig. 1, is such that the flow is
fully 3-D. Due to the axial symmetry of the tank, it is possible to avoid to simulate the
transient flow induced by the rotation of the impellers. We can simulate the flow in a rotating
frame of reference, by adding or subtracting an angular velocity . In this case, the flow is
steady. As we are interested by inertia effects on the flow characteristics, the centrifugal and
Coriolis forces are added as body forces.
Let us consider the following mixing flow process: a fluid is contained into a cylindrical
vessel, and is set into motion by means of two blades. The angular velocity of the blades is
50 rad/sec. The radius of the cylindrical vessel is 10.5 cm and its height is 24 cm. Those
blades are attached to a central axis. This description corresponds to what is seen from a
fixed observer. Gravity effects are taken into account. The tank is not closed: the upper wall
is removed and the fluid is in contact with the air.
In a rotating frame of reference, the observer sees fixed impellers, the wall of the vessel being
rotating in the opposite direction. Since inertia is taken into account, centrifugal and Coriolis
forces are added into the momentum equation.
The angular velocity at the wall will be 50 rad/sec. We will apply an evolution strategy on
the fluid density in order to analyze inertia effects as the Reynolds number increases.
KEYWORDS
moving frame of reference, rigid rotation (centrifugal and Coriolis forces), 3-D flow,
evolution, sub-models, mini-element
FILENAMES
Directory: stirring_tank
SYSTEM OF UNITS
MESH GENERATION
The finite element mesh contains five sub-domains. The internal blades are not boundaries of
the flow domain, but intersections of sub-domains (see "new concept" paragraph for more
information). The boundary 1 is the bottom of the tank, the boundary 2 is the lateral wall of
the tank, the boundary 3 is a "free surface" and the boundary 4 is the internal axis to which
the blades are connected.
Sub-models allow you to add additional constraints on variables (velocity, temperature, force)
on every part of a mesh that can be defined as a combination of unions and intersections of
existing mesh parts (sub-domains, boundaries and P-meshes) (see also note 5 below). In
Polydata, we can define such new parts and impose constraints on them. This is done in the
option "Define sub-models" in the menu of the F.E.M task. Let us note that there is no check
of the validity of such constraints.
In this example, as the blades are very thin, we simplify the mesh generation by neglecting
their thickness: we define them as an intersection of some sub-domains. In Polydata, we
define such a 'topo-object', and impose a vanishing velocity on it. The assumption of zero
thickness for the blades will limit the choice of interpolation for the velocity and pressure
fields (see also note 6 below).
POLYDATA SESSION
The task set-up is also set to "rigid rotation". This means that centrifugal and Coriolis forces
will be added for all sub-tasks. One additional parameter is the angular velocity of the
rotating reference frame (or the rotating observer) with respect to an inertial reference frame.
We select this option, since our flow involves inertia; moreover, we will apply an evolution
scheme on this material parameter. The fluid density will increase with the independent
evolution parameter S. The parameters for evolution (initial S, final S and S-interval are
specified in the 'Numerical parameters').
The problem is described with respect to a rotating reference frame. The observer (on the
blades) thus sees a rotating vessel. Here we must specify the rotational velocity of the
boundary of the vessel, with respect to the observer.
The tank does not have an upper wall closing it: the fluid is in contact with the air. Ideally,
we would have to calculate a flow with a free surface, leading to additional numerical
complexities (free surface involves remeshing algorithm). However, in this case, we are not
interested by the shape of the free surface. Consequently, we choose to replace a free surface
condition by a "full slip condition", which neglects any motion of material points
perpendicularly to the surface.
Note 5: Interpolation
With sub-models, we must choose carefully the interpolation for velocities and pressure. In
this peculiar case, the pressure must be different -a priori- on the two sides of the blades:
consequently, a linear pressure can not be chosen for calculation, because this lead to the
same pressure on both sides of the blades. In this example, we chose the mini-element for
velocities and a constant pressure per element. However, let us remember that in Fluent-Post,
all the fields are linearly interpolated: what you see is not unfortunately- what you
calculated !
We can combine any objects of this list to calculate a new object with two operators:
intersection and union. Recursively, we can use objects created by the user to define other
ones. We can not combine objects having different geometric dimensions (example: to obtain
the union of a point and a surface). Moreover, if the result of an operation is "empty", an
error message is sent to the user and the resulting object is not created.
RUNNING POLYFLOW
The input files for POLYFLOW are stirring_tank.msh (the mesh) and stirring_tank.dat (the
data file generated by POLYDATA).
GRAPHIC POST-PROCESSING
In Fig. 2, we display two typical iso-surfaces of the vertical velocity obtained at various steps.
We observe that the intensity of re-circulation increases with the Reynolds number: the green
volume is the fraction of fluid having a vertical velocity lower than 1.0 m/s (fluid moving
down the tank); the brown volume is the fraction of fluid having a vertical velocity greater
than 0.5 m/s (fluid moving up the tank). The mixing should be better at high Reynolds
number (let us remember that we are still in a laminar regime).
a) b) c)
Fig. 2. Isosurfaces of the vertical velocity for various S-steps: a) step 1, density = 0,
b) step 3, density = 337.5, c) step 6, density = 900. Green iso-surface corresponds
to Vz = -1.0, Brown iso-surface corresponds to Vz = 0.5.
In Fig. 3, the pressure field is seen for the same steps as in Fig. 2: we observe that the pressure
variations increase with the Reynolds number, but also that, at high Reynolds, the pressure
increases from the axis to the tank lateral wall in a radial way.
a) b) c)