Method Development for CFD Analysis of Heavy Truck Aerodynamics: Investigations of Effects of Partitioning for Parallel Simulations, Bi-stable Flow States for Steady Simulations, and Effects of Split-line Variation on Drag
Typ
Examensarbete fรถr masterexamen
Program
Applied mechanics (MPAME), MSc
Publicerad
2020
Fรถrfattare
Larsson, Henrik
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
According to the European Commission, heavy trucks are responsible for about 4% of total European CO2
emissions. One way to lower emissions is to reduce aerodynamic drag. In order to effectively design low drag
trucks, computational
fluid dynamics (CFD) simulations are used for the analysis of aerodynamic performance.
This thesis consists of three parts that investigate CFD simulations for truck aerodynamics in three areas.
Firstly, differences appearing between parallel simulations are investigated. As truck aerodynamics is very
computationally expensive, simulations are run on parallel computer clusters. It has been seen that running
simulations on different clusters may cause differences in simulation results, which should not should not
occur according to theory. In the second part two bi-stable
ow states that appear in steady simulations are
investigated. Small geometrical variations cause the solution to enter one of two modes, with no clear pattern.
Lastly, real trucks will deviate from the nominal geometry to some degree, due to natural variation. Simulations
are however mostly performed on nominal geometry, which leaves some uncertainty regarding the performance
of real vehicles. Therefore the effects of split-line variation on drag are investigated.
A design of experiment (DOE) approach was used to identify which parallel parameters were affecting results,
both in parallel meshing and parallel simulations. It was found that parallel effects on the meshing was
negligible. The differences were found to appear when running simulations on different clusters and different
number of processors. The cause for the difference was an underlying instability in the form of the bi-stable
ow states. Differences in the algebraic multigrid (AMG) linear solver or in
floating point errors caused
sensitive simulations to change mode. Differences between clusters could be eliminated using the command
-mppflags "-e MPI_COLL_FORCE_CONSTANTORDER=1" in Star-CCM+ for clusters with Platform MPI (Message
Passing Interface).
The bi-stable modes were investigated for some different cases. The
flow fields, solution development, local
convergence and streamwise pressure gradient were investigated. A mesh study was performed, and a
turbulence model study with both Reynolds-averaged Navier-Stokes (RANS) models and a detached eddy
simulation (DES) model. The modes were partially explained by interaction with the front door split-line.
One of the modes, Mode 1, was found to be most probably physical. The physicality of the other mode
could not be determined with certainty. It was however found not to be the dominant mode. A possible
solution to the problem is to use the realizable k ๔ " model, which was most consistent with the modes predicted
by the DES. Alternatively a more accurate transient model such as DES can be used, if it can be afforded.
Split-lines in four different areas of the truck cab were investigated using a design of experiment approach.
Gap and
ush at minimum and maximum tolerance values were considered for the split-lines of the selected
areas. It was found that the impact on drag is within 5 DC (drag counts), and that most trucks will be
within 3 DC of nominal CD. The most important of the investigated areas are the front corners corners of the
cab. It was furthermore found that gap has a larger effect on drag than
flush, for the split-lines and tolerances
considered.
Beskrivning
รmne/nyckelord
CFD , Commercial Vehicles , Aerodynamics