| |
By Tiberiu Barbat, Fluent Inc.
View the pdf of this article
Anyone who has driven a car with a nearly empty gas tank
knows the sinking feeling that sometimes occurs when the car
accelerates. Depending on the design of the gas tank, the gas
gauge can register below its actual value during acceleration, giving
the driver some anxious moments that may not be justified. To investigate
this effect, the volume of fluid (VOF) model in FLUENT has
been used to study liquid sloshing in two gas tanks during acceleration.
One of the tanks is fitted with baffles to help stabilize the liquid
motion, and the other is not. Both tanks have a fuel pump intake pipe
that draws fuel from the tank when the engine is running. The goal
of the simulation is to determine which design provides an uninterrupted
supply of fuel to the intake pipe during vehicle acceleration.
The location of the liquid fuel and the velocity field on the free surface for the
unbaffled tank design after 1 second, solved using NITA; results showed that the
fuel tank intake was out of the liquid at certain times during the simulation
The liquid fuel free surface and velocity field for the baffled tank after 1 second,
solved using NITA; the design enforces recirculation between the baffles, limiting
the amplitude of the sloshing waves, stabilizing the fuel, and keeping the fuel
pump intake submerged
The car is assumed to accelerate in a horizontal direction for 1 second
at 9.81 m/s2, corresponding to a practical situation of accelerating
from 0 to 22 mph in 1 second. The liquid fuel in the tank (with a
total capacity of about 75 liters or 19.4 US gallons) is initially at rest,
with its surface 8 cm above the bottom of the tank. At this level, the
tank is one-quarter full. The geometry and hybrid mesh were created
using GAMBIT. Regions of hexahedral mesh obtained using the versatile
Cooper meshing algorithm were easily combined with a zone
of tetrahedral elements in the region of the fuel pump pick-up pipe.
In FLUENT, selected face zones were set to walls for the case of the
baffled tank, and set to interior faces for the case of the unbaffled
tank. Thus, the same initial mesh file could be used to simulate the
flow inside the two tank designs considered.
The VOF model used for the simulations is designed for applications
involving two or more immiscible fluids. A single set of momentum
and continuity equations is solved, but different property sets are
defined for each fluid. The volume fraction of one phase (in this case,
the liquid fuel) is tracked throughout the solution to determine which
fluid occupies each computational cell at any given time. In cells containing
both fluids, a special routine is used to predict the shape and
location of the interface. For time-dependent simulations such as the
one described here, a special algorithm is used to update the volume
fraction in a cell from one time step to the next, based on the velocity
field delivered by the solver. The geo-reconstruct algorithm used
for this example is capable of capturing a sharp interface between the
phases. The time step size used for most of the simulation is 2.5 milliseconds,
so 400 time steps were required to model 1 second of real
time.
Comparison of a typical problem variable, the turbulent kinetic energy, at a
monitor point inside the liquid for the baffled case using FLUENT 6.2 with
NITA (red) and FLUENT 6.1 (blue); the time to complete the simulation
using FLUENT 6.2 was 8.5 times faster than when the iterative approach
in FLUENT 6.1 was used
The problem was solved in FLUENT 6.1, and comparatively in
FLUENT 6.2 with the new non-iterative time advancement (NITA)
algorithm. This new algorithm, which requires only one global iteration
per timestep, is compatible with the VOF model. For the baffled
case, it led to a speed-up of about 8.5 when compared with FLUENT
6.1. Predictions of surface patterns from FLUENT 6.1 and FLUENT 6.2
were indistinguishable.
|
|
|
