You are on page 1of 7

HIP-19-036

Application of Helios to the Workshop for Integrated Propeller Prediction


Aaron Crawford1,2, Mark Potsdam3

1 HPC Internship Program, DoD High Performance Modernization Program


2 University of Tennessee at Chattanooga
3 US Army Aviation Development Directorate, Moffett Field, CA

Abstract
Initiated in Summer 2018, the AIAA began the Workshop for Integrated Propeller Prediction
(WIPP). Through this program, using the Maxwell x57 as a test bed, the goal was to develop the
CFD prediction capabilities for integrated propeller models. The reason behind this workshop
was to develop prediction capabilities due to the lack of available and non-proprietary CFD
research in this area, as well as an attempt to keep up with the renewed interest in propellers for
electric applications. My work began at the University of Tennessee at Chattanooga in
September 2018, generating meshes and running simulations for the proposed wind tunnel test
conditions. This project was continued during the summer of 2019 at the US Army Aviation
Development Directorate office at the NASA Ames site in Moffett Field, California. Using the
tools from the CREATE-AV suite including Helios further CFD simulations were completed.
One set of run conditions was applied to several geometry configurations: an isolated propeller
and nacelle, an Isolated wing, and a full configuration. Additionally, a traditional near body
solver, FUN3D, was used alongside the newer strand solver unique to Helios, mStrand. Though a
full configuration mStrand solver was never completed, both solvers performed well on this case.
For each geometric set up, both solvers predicted workshop wind tunnel test values. This
includes lift, drag, thrust, and Cp along the wing upper and lower surfaces. All values were
predicted well except for drag. Despite the poor prediction of drag, all configurations and
solvers performed well with this research, helping lay some groundwork for the CFD prediction
capabilities of propeller driven aircraft.

Introduction
The focus of this research is a workshop hosted by AIAA at
their annual Aviation conference. The goal of the
Workshop for Integrated Propeller Prediction (WIPP) is to
develop the overall CFD prediction capabilities of the
aviation community, particularly regarding propeller driven
aircraft. This stems from a renewed interest in propeller
driven aircraft for the purposes of electric flight, and a lack
of nonproprietary data about propeller aircraft. In this case
the workshop consisted of blind CFD validation of
experimental data. The test bed for this workshop was a 0.4
scale model of the x57 Maxwell, the NASA developed
electric propeller driven aircraft. The wind tunnel tests
occurred in January 2019, with the first workshop meeting
in June 2019, to be followed hopefully by a second
workshop in June 2020. Simulations were to be completed
by various AIAA members and presented blindly at the first
workshop.

Page | 1
HIP-19-036

Initial Work
I became involved with the workshop in September 2018 at my home university. At the
beginning of the workshop, no wind tunnel tests had been performed, so I was given the task of
generating meshes from the given models. These models were CAD representations. Early in
2019, the wind tunnel tests occurred, and the full test CAD model was released. This began
another round of meshing which now included the propellers. Also included in the
computational mesh was a sliding interface. The sliding interface allows for the rotation of the
propeller and nose inside of a domed volume while the rest of the model stay stationary.
Unfortunately, due to the lack of available wind tunnel data and unrefined mesh issues, the
results produced were not as accurate as hoped for. Fortunately, the background, experience
with the project, and the lessons learned allowed the work in the summer to progress further than
it would have otherwise.

Work with Helios


At the start of the summer, the initial goal of this project was to continue research on the WIPP
project and accomplish as much as I could, since simulating and processing data from the whole
workshop would be unfeasible. Also, for the first three weeks, no work could begin since the
wind tunnel data from the workshop was not released until it was held in mid-June. The first real
step to getting started on the workshop started with attending the workshop.

While attending the Aviation conference workshop, I was able to hear from others who had been
involved with the workshop, those that did the wind tunnel testing, and those who did
preliminary simulations before me. Once back and working with Helios, I first needed to adapt
my initial surface grids and volume meshes. The first change was to alter the geometry so that
there is a gap between the propeller and hub and the nacelle. This is necessary because Helios

can handle relative motion without a sliding interface. Unlike many others at the workshop, I
chose to run these simulations with the full model blades spinning at the correct RPM instead of
using a less computationally expensive lower order actuator disk model. I chose to run cases with
both FUN3D and mStrand as both a test of the two, but also as double verification of
computational results. mStrand is an exciting new strand solver that is still being developed and
this project seemed like a great opportunity to test its capabilities. For the Fun3D cases, the gap
before the nacelle was all that was needed. For the mStrand cases, further changes needed to be
made. For these cases, the Hub was shrunk slightly so that it was separated from the propellers.

Page | 2
HIP-19-036

Simulation Setup
I chose to focus on 1 test Table 1: Test Conditions
condition, but test multiple
configurations with both near Freestream Mach 0.11
body solvers. Angle of Attack 0̊
The three geometry RPM (Prop Cases 8060
configurations that were tested Only)
were prop hub and nacelle, Time Steps/ Iterations 5 Revs (7200 steps at ¼ rev/step)
Isolated Wing, and the full 10000 Iterations
testing configuration. Table 1
shows the test conditions of run 03, and 176.

I encountered several problems along the way.


Firstly, with the FUN3D case, the gap between
the nose and nacelle developed regions of high
and low pressure that should not occur and
caused high fluctuations in thrust, despite the
residuals from these cases being converged. This
is shown to the right. To combat this issue,
FUN3D inputs as well as pundit inputs were
adjusted, specifically the flux limiter and the m-
exclude values. Also, in FUN3D, there was an
issue of mismatched boundary conditions at the
base, causing high pressure and vorticity. This
issue has not fully resolved, although changes to
the FUN3D boundary conditions has improved
this issue.

The computations for this project were performed using DOD supercomputing facilities,
primarily the Navy supercomputer Conrad. The supercomputer housed at the ADD office,
Blackhawk, was also used primarily for smaller jobs and debugging. Typically, simulations
could run anywhere from one day up to five occupying 16 nodes and 512 cores.

Results
For the test condition outlined above, successful simulations were run for each of the geometry
configurations except for the full mStrand case. For all the completed cases, both near body
solvers predicted thrust and lift well within 5% of workshop test data. Drag is unfortunately a
different story. For both solvers, the CD calculated was approximately 40% off with the
mStrand solution being the closest at 33% error. With the full configuration model, this was not a
problem since in that configuration the CD contains the thrust which dominates making the CD
highly accurate. Moving forward, I intend to correct this and produce results for drag that are
more reasonable.

Page | 3
HIP-19-036

Table 2 Experiment mStrand Fun3D


Run 03, 176
Propeller and N/A 151.16 151.72
Nacelle Thrust

Isolated Wing CL 0.5806 0.577 0.6005


Isolated Wing CD 0.0302 0.0403 0.0423
Full Model CL 0.7012 N/A 0.6734
Full Model CD -0.3757 N/A -0.3755
Thrust Included
Full Model CD 0.0255 N/A 0.0404
Thrust Removed

In addition to the force measurements, the workshop included the pressure taps along the
wing chord at different points along the span. All 6 pressure lines were simulated for both
isolated wing cases and the full model FUN3D case. For all these pressure lines the simulations
predicted Cp very well. Below is a graph of both the mStrand and FUN3D solutions in line with
the experimental data. As is visible both mStrand and FUN3D solutions are nearly identical and
line up well with experimental data. This accuracy also applies for the FUN3D full
configuration, particularly the Cp pressure lines directly in the downwash of the propeller. The
two discontinuities that appear in the Cp plot are due to actual physical deformities that occur in
the model. Since the wind tunnel model was scanned, the final simulation model contains all of
the true to test inconsistencies that come with any manufacturing process.

The workshop also produced wake


data. From this wake data, torque and
swirl can be calculated at multiple
locations downstream as well as in the
radial direction. Unfortunately, at this
time, this computational data has not
been processed in order to compare
with the experimental results. However,
it is suspected that these simulations
will perform well, given that no lower
order models were used and that for the simulation the blades were fully modeled and spinning at
the correct rpm.

A result that should be further explored is the interactional effect of the propeller on the
wing and the wing on the propeller. The graph below shows the oscillation in lift and drag on
the wing. These oscillations occur as a result of the vortices from each propeller hitting the wing.
As is visible over the course of one revolution, there are four almost identical cycles in lift that
occur. A little less noticeable are the cycles in drag. Because of the smaller scale of drag

Page | 4
HIP-19-036

oscillations, it is harder to notice, but in a similar manner the maximum and minimum line up
almost exactly with those of lift. Further studies into the spanwise distribution of the force from
these vortices could be examined. Also, looking into exactly when each vortex arrives at the
wing and how it interacts with the top and bottom wing surface would give a better picture of the
interactional effects at work in this model.

When comparing the performance of each near body solver, both perform well. In comparing
thrust for the isolated propeller case, lift and drag for the Isolated wing case, both solvers
predicted these values within 5% of each other. The slight edge in accuracy is given to mStrand
which narrowly surpassed FUN3D with slightly more accurate predictions. Additionally, as seen
in the plot below, mStrand seems to converge slightly faster. Overall the performance of both
solvers was satisfactory, and the narrow difference between them shows the promise that
mStrand exhibits. No computational comparison can be directly made between these solvers
though. They ran under different conditions with different numbers of points as well as different
numbers of sub-iterations. However, a direct computational cost-effectiveness study could be
done on this case.

Conclusion
Overall this project was
successful in the simulation
of a set of test conditions for
the WIPP using Helios. In
addition to the test
conditions, it was also
successful in comparing
these cases across two near
body solvers and three
different geometry
configurations. There is significant room for improvement in this project. Despite the need for
improvement, the summer was a success in laying the foundation for the work that will be
performed this following year. Hopefully, the troubles encountered thus far mean that continued
work will continue more smoothly.

Page | 5
HIP-19-036

Future Work
In the future I plan to continue this project. At my home university, we are working to get
Helios as well as FUN3D so that I can continue this work. In the short term I will continue to
analyze computational wake data and compare this with wind tunnel wake data. Additionally, I
will be working to create a working full model case using mStrand. Transition and turbulence
models are things that have not been considered up to this point; however, they will be in hopes
of correcting the large error in drag. In the time since completing the internship, this has already
been implemented with promising results, reducing the error in drag by a factor of four. This
work will be the subject of my senior thesis as I complete my bachelor’s degree. This thesis
involves continued work on the project over the next few semesters, ending in a defense. I will
continue to simulate the rest of the run conditions provided by the workshop. These conditions
include varying the angle of attack, rpm, and freestream mach. I will also examine the wake data
from simulations and further explore the propeller wing interactional effects. I plan to present
this research at the 2020 aviation conference, either in the workshop or as a student paper.

Impact
For me this work has a substantial impact. While
my university has several great CFD, faculty it is
not a large topic especially since it is under the
context of mechanical engineering. This summer
was my first real experience in a fully aerospace
environment. Over the summer, I developed skills in
several pieces of software and programs including
Helios and Higen, FieldView, Linux Shells and the
DOD supercomputing facilities. In addition, the
summer included much learning through simply
interacting with the ADD employees, as well as
taking part in many of the NASA sponsored tours
and events. The NASA Simulation Seminars and the
NASA summer series talks were all incredible and
particularly informative. Over the 10 weeks, I was
able to do things I never thought that I would,
including tour both a C130, seen to the side and a
pavehawk, see the cross section and drivetrain of the
world’s largest wind tunnel, and earn the game ball
from a NASA league softball game. The internship was eye opening and helped confirm that
aerospace and CFD are the fields I would like to continue to pursue into a Ph.D.. It is still too
early to say where or what graduate school will look like for me; however, this internship has
encouraged me to go and continue developing the skills and knowledge learned in the ADD
office. In addition to the academic opportunity, getting to travel to California was a wonderful
experience that might never have happened were it not for this internship.

With an interest in green and electric aerospace vehicles, propellers are becoming more
prominent in the aerospace landscape. This research and its continuation help to lay the
groundwork in several important areas. This research helps build the general body of knowledge
surrounding propeller CFD predictions. It also serves to advance the application capabilities of

Page | 6
HIP-19-036

new technologies like mStrand. With the shift toward green technology, propellers will be more
important than ever especially when considering EVTOL and urban air mobility. The more we
understand about propeller technology and the better we can predict it through more advanced
and less expensive simulations, the more we can improve overall usability, reliability, and safety
of the next generation of aerospace products.

Acknowledgements
I would like to acknowledge the entire ADD team for hosting and helping me through this
project. Particularly Mark Potsdam, my mentor for the summer. He endured a lot of questions,
and I would not have made it as far in this project without his help. I would also like to
acknowledge the support of Buvana Jayaraman, Vinod Lakshminarayan, Jared Sagaga and
Austin Thai. Each one of them helps substantially, whether it was with starting paperwork, many
mStrand or FUN3D questions or anything else related to the CFD process. Lastly, I would like
to thank the DOD HIP program for allowing me this opportunity, and giving me access to the
supercomputing facilities that make this work possible.

Page | 7

You might also like