For the numerical analysis, an open source Lattice BoltzmannMethod (LBM) tool called OpenLB is used. The tool combines the Boltzmann equation together with Large-Eddy Simulation (LES), as turbulence model, to simulate the fluid behavior. The combination of LBM-LES which is used by OpenLB is relatively new within the CFD realm. To asses, the performance of OpenLB in the field of flow characteristics, drag prediction and pressure patterns a generic heavy duty bluff vehicle called, Generalised European Transport Model (GETS), is used. The inherent unsteady nature of both LBM and LES are complementary to each other, where LES is accurate in prediction of vortex shedding behind bluff bodies and with the easy parallelization of LBM
computational time can be saved. A bluff vehicle such as the GETS model, where separation is expected, is the
perfect model to test the performance of OpenLB and compare it with established phenomena of bluff bodies
in literature and wind tunnel tests. By changing the configuration of the GETS model, variation in frontedge
radii and additional boat tails, a better insight into emerging trends can be observed and compared,
to validate the open source CFD tool. Furthermore, to investigate if there is potential in OpenLB for future
projects.
In addition to numerical analysis and a wind tunnel experiment is performed to obtain drag coefficients
for the various model configurations. The experimental analysis was performed at Reynolds number varying
from 8000 to 60000, based in the square root of the frontal surface. This means that a 1:50 scale model of the
original GETS model is used. Mainly, the drag force of themodelwith different configurations are measured to
validate the numerical results. The relative low Reynolds number tested in this study is due to the limitations
set byOpenLB. The absence of grid refinement, the use of LES as turbulencemodel and lack of a wall function,
constrains the Reynold numbers that can be simulated with keeping computational time in mind. Therefore
a choice was made to only simulate the Reynolds number of 8000, 24000 and 48000 and compare the effect it
has on the flow characterizer, drag coefficient and pressure.
It was shown that the model with the smallest front-edge radius had the highest drag coefficient. This was
the effect of a separation bubble over the front of the model. With increasing the radius the drag is also reduced,
observed both by numerical analysis and conducted experiments in literature. The flow is guided over
the rounding which gives a more favorable pressure gradient and therefore reduces the separation bubble
that arises over the front-edge. Additionally, the effect of applying a simple Bounce-Back or Bouzidi boundary
condition on the model is investigated. The interpolative nature of Bouzidi approximates the staircase
shape of the rounding with a curve, in contrary to the bounce-back boundary condition. The drag difference
between the two boundary conditions was 7.3%, which is moderate but it highly affects the flow over
the front. The flow characteristics are simulated at two different heights, one at ground proximity and the
other at a higher distance fromthe ground. The comparison is made because with the experimental analysis
the model was placed somewhat higher from the wind tunnel floor to avoid interference with thewind tunnel
boundary layer. Normally themodel is situated at the ride height of a real-life truck. The position of themodel
influences the location of unequal sized vortices aft of the model. At ground proximity, the largest vortex is
at the top whereas at a higher ride height this is vice versa. Which complies with Particle Image Velocimetry
(PIV) performed by van Raemdonck [78]. The ride height did slightly affect the drag coefficient but was
within a reasonable difference of less than 3%. Another observation that is made is, with increasing Reynolds
the drag coefficient decreases. This is valid for all the front-edge radii for both numerical and experimental
analysis.
The addition of the tail lowered the drag of the models, this trend is ascertained by both numerical and
experimental analysis. With increasing tail angle the drag reduction also increaseswhich is caused by increasing
pressure over the rear part of themodel. There are some exceptions. Experimental results show that with
the largest tail deflection of 18± the drag is increased. This can be the influence of how the tail is attached to
themodel and the material used to create the model. A CADmodel that is used in the numerical simulation,
every angle and dimension is perfect to the specifications. However, with wind tunnel models this not the
case and therefore has an influence on the results. In general, the numerical and experimental drag difference
is rather high, varying from 50% at low Reynolds numbers to 10% at higher Reynolds numbers. This is
mainly the cause of the force balance used to measure the forces in the wind tunnel. The balance is designed for larger models at higher inlet velocities which generate a larger force. So, at smaller forces, the balance is
not that sensitive which gives a wider spread in results at lower Reynolds numbers.
Comparing the flow characteristics of the model with the different configurations a few things can be
noticed. That the separation bubble is reduced with increasing radius. Also, the strength of the recirculation
region of the bubble is reduced. This is one of the reasons why the drag coefficient is reduced with increasing
front-edge radius. Furthermore, the effect of the additional tail on the separation bubble is also visible. The
flow over the rear is accelerated by the tilted plates of the tail, which affects the boundary layer over the entire
model, it re-energizes regions of low-velocity flow reducing the size of the separation bubble. Comparing
the simulated flow behavior of boat tails with a slant angle of 6± 12± and 18± to literature shows satisfactory
results. The addition of the tail has three major effects on the flow aft of the tail. First, reduced the wake size.
Second, pushed the wake more aft so it has less influence in the model and third, delaying separation over
the tail and guiding flow more inward to reduce the vortex strength. All of these core functions of the boat tail
are simulated correctly. At each simulated angle the flow characteristic match that described in literature.
The pressure coefficient could not be compared because the position of measurement was not equal.
Due to an absence of a wall function the pressure at the wall of the OpenLB simulation are zero therefore no
true comparison with literature can be made. In literature the pressure is often measured at the face of the
body. Hence, no qualitative comparison could be made about the magnitude and the shape of the pressure
distribution. However, certain trends are visible with the pressure plots. The larger boundary layer that is
caused by a sharp front-edge radius decreases the pressure over the aft of the model, whereas, with a larger
front rounding the pressure is increased. The addition of a tail clearly increases the base drag and therefore
reduces the drag contribution. The pumping effect, that is very common with bluff bodies, is also observed
with the help of the pressure plot. The pumping effect is, in fact, a periodic motion that sheds rear-end
vortices generating a longitudinal oscillatory motion. This is reflected in the pressure coefficient plot which
shows the back and forth motion with increasing and decreasing pressure coefficient.
In conclusion, it can be stated that OpenLB is not yet ready to be used for the mainstream engineering
problems. The absence of certain key features limits the tool in many ways. Using LES without grid refinement
and wall function only low Reynolds numbers can be simulated. Overall, the flow characteristics that
are simulated with the various model configurations are accurate even if the Reynolds numbers are not the
same order of magnitude. The drag prediction of the simulations are underestimated if compared to the experimental
results. This could be due to the choice of the LES model or the effect of too much dissipation
which reduced the drag coefficient. The force balance used in the wind tunnel has also a major influence
on the results, by being not sensitive enough at lower Reynolds numbers. Although the pressure coefficient
did not match with that of literature some important bluff body phenomenon could be observed from the
trends.