Day 1 winners Japan Team SF Bay |
The panel: Mike Drummond, Technical Director, SailGP, Edwin Upson, GVP,
Enterprise Cloud Architects, Oracle, Warren Jones, Director of Technology, SailGP,
Hans Henken, Flight Controller USA SailGP Team, Tom Burnham, Coach, SailGP and Phil
Crane, Data Analyst, SailGP.
WHAT ITS LIKE FOILING A F50: It's always appreciated when techies can bring the language
to the layman. Mike Drummond, describing sailing a F50 catamaran up on its
foils, said, “It’s like balancing a broomstick on your fingertip. Except this broomstick on your finger is
actually only a few inches long, and has another
full-length broomstick, with a broom on the upper end of it, way up in the air,
balanced on top of it. And it’s not
exactly on your fingertip; it’s on the fingertip of your friend, standing right
next to you, and you are telling him how to move his finger to keep it balanced
there.”
To keep these boats sailing on their foils, the sailors are
constantly adjusting the angles of the foils and sails to achieve and maintain
a most opportune “ride height” of the leeward hull above the water.
Very small
changes to the bow-up attitude make a tremendous amount of difference: Russell
Coutts indicated during a boat tour that a difference of 50cm in ride height
when foiling can mean a difference of two to four knots of boat speed. Too low, and your hull slaps the water and
that’s very slow. Too high, and the water foil will cavitate (the water around
the foil breaks down / boils / evaporates), and you crash back down into the
water while going 25 mph.
Hans Henken |
Henken continued, “Pitch and ride height are the comms
between the flight controller and the helmsman. More bow down attitude on the
boat provides more speed. Our wing trimmer Riley Gibbs changes the attitude and
twist of the wing in order to give us the power we want, when we want it.”
Getting and keeping these things flying is tremendously hard
work, and it is difficult. Data has
helped these sailors to better learn and understand their flying machines. The
boats are highly instrumented, with1200 sensors onboard each vessel. Oracle is
helping SailGP to collect, display, transmit, store, and analyze 20 MB of data
from every race boat, per 15-minute race. With 1000 channels of information
from sensors onboard, and with some channels operating at 50 hz (readings per
second), it’s a significant data stream. From it, they are inferring the behaviors of the boat.
HOW THE DATA FLOWS: SailGP proprietary IP is layered atop an Oracle open
platform. Oracle intends that model to be replicated for additional sports
applications. The SailGP electronics team build some of their own gear, using
some conventional marine electronics sensors (from B&G and others), feeding
real-time info to on-board displays for vital boat performance and control
information.
Every race boat has individual SIM cards, hubs, and their
own bandwidth, and their performance data is relayed in real-time via telemetry
across a low-latency, custom-built LTE (4G wireless) network built with
multiple antennas around and upon the waterfront race area.
The data (and audio, and video) collected on each race boat
flows into a popup data center and into the cloud, at 200 ms. Live video
screens onshore, on coach boats, course marshal boats and other displays are
powered from a 100 ms from SailGP in London. All of this info mashes up into
the SailGP app that you was available throughout last week for practice races and
during racing on the Apple Appstore (the app was released to Android Thursday
May 2).
HOW THE DATA GETS USED AFTER SAILING:
Data gets loaded into autonomous data platforms that include
AI. Oracle engineers can set up the data they want to look at, like a Spotify
playlist. Teams have hired data analysts to interpret and help teams to learn
from data collected and uploaded from sailing sessions and races to better inform
performance.
Coaches Analyse Racing |
Before San Francisco , SailGP
did not have an adequate dataset to fully analyze the performance of these
newly-configured boats; conditions at the Sydney
event had light breezes, and they may not have had the opportunity to collect
enough race-quality data in heavier air conditions. That will not be the case
after the end of this weekend’s racing; the dataset will be more robust, well-developed
and useful, and can be compared meaningfully against designer’s polars and
targets.
China Team Day 1 SF Bay |
DATA IS SHARED AND OPEN TO ALL TEAMS:
Unlike most other types of sailing, each boat gets to look
at data from all the other boats, in real time and during post-race analysis. This
is in marked contrast to the Americas Cup, and almost all other sailing.
Teams that weren’t performing well in Sydney have greatly
improved here in San Francisco in the last week, by using stored performance
data, and comparing their sailing efforts to those of the teams who made their
boats perform better. In the future, new teams will be able to rapidly get up
to speed by learning the boats virtually at first, with stored data, and then
going sailing and trying to emulate bench-marked performance achieved by
experienced and well-practiced teams.
The data being collected is leading directly to design improvements,
including larger wings for lighter venues, and possibly different rudders. The
boats will remain one-design but will evolve. Specific suggestions by
individual teams are considered for applicability for inclusion in design mods
to all boats. There’s a lot more technological development continuing upon the
racing boats, and while the data being used for boat technology development and
ability to fly them is the most interesting to some of us, SailGP is focused on
tech development well beyond the race boats. “We are optimizing everything in conjunction…boat tech,
broadcast tech, fan experience, and it’s a lot!” Drummond noted.”
Japan Gybes Day 1 SF Bay |
Tom Burnham, SailGP coach, and Phil Crane, SailGP Data
Analyst, worked together at the last Americas Cup and it seemed natural to do
it again together at SailGP. Beginning
at the first event in Sydney, Phil prepared numerous reports based upon
extensive data analysis. They are working now in San Francisco at doing this
even better, and they are working on building their abilities to do more meaningful
and effective real-time data and performance analysis – and providing feedback,
based upon that data, to the teams, both in
real-time and in post-race debriefs.
This real-time feedback and input from the coach to the
teams is enabling improvement during the
day while out sailing and racing, building significantly upon the
effectiveness of traditional post-race analysis. In conventional sailing, this
is prohibited by the racing rules, which prohibit communications or access to
information that is not specifically and freely available to all competitors.
SUPER-DETAILED POST-SESSION ANALYSIS: After each sailing day, there is typically a 1-hr debrief
post-race with the coaches and (probably very tired) sailors. They go through
race videos, audio, hand written notes, audio notes, and the post-race debrief generates
LOTS OF QUESTIONS from the sailors who, during the race, had their hands full
just trying to keep the boat flying. When the action is over, they want to know
how to fly better.
Tech Services F50 Foil |
Boat sensor data, coupled with onboard video, onboard audio,
and all boat’s GPS positions displayed in the most appropriate ways enable
immersive re-creation of a session or race for analysis. More details provide
better data analysis, and a more realistic and accurate recreation may enable
noticing small things that might have had unanticipated or surprisingly large
inputs into a process - whether it might be a crash or a particularly nice
liftoff, tack, or gybe.
After the debrief, Crane digs more deeply
into the data, sometimes can answer some of the ‘lots of questions’, and provides
detailed reports. Comparing actual performance achieved in a race to theoretical
models, benchmarked best behaviors, speeds and angles, and to performance of
other teams sailing in the same conditions, in that same race, enables detailed
and quick learning opportunities. Teams can look at all of the settings they
used onboard for sail and foil controls, compare them to the settings used by
other boats in that race and others, and rapidly learn to do things better.
Many F50 settings are automated within control buttons used
on the systems. Super-detailed data logging enables post-race analysis of exactly
when buttons were pressed, in what sequence, and by whom, to minutely examine
what works, and what doesn’t work, as these talented athletes learn to fly
these sailing machines.
With continuous improvement of systems, sails, foils,
controls, and the sequencing of what they do with all of them, these guys are
going to soon learn to balance their broomsticks and foil all the way around
the race courses. A race with multiple boats that never slow below 19 knots
will be a whole new and exciting kind of simultaneous and competitive broomstick-balancing.
F50 Performance data
seen on screens during Thursday May 2 practice race, with winds 15-18 knots:
Bottom speed in a foiling tack: 18 kts
Takeoff speed:19 kts
Upwind boat speed of 27 kts, VMG (velocity made good towards
the wind) of 18.3 kts
28 kts target boatspeed upwind
40 kts of boatspeed achieved at bear away at top mark
23 kts: Lowest boatspeed during a foiling gybe
Note: Thanks to John Bonds for providing this report
Photo Credit: SailGP
Photo Credit: SailGP