These web pages describe a
project to build a low cost device that records the stress on an airplane. The data from such a device can be used for engineering
analysis and for making operational decisions such as when to inspect, repair,
rebuild, or retire aircraft structures.
The inspiration for this project comes from the FAA’s intention to issue
rules regarding when twin Cessna wing spars must be rebuilt (see http://twin.cessna.org/ and http://twin.cessna.org/referen
The current FAA direction is
to require the wing spars of some models of twin Cessnas to be rebuilt after a
certain number of flight hours. This
rebuilding is an expensive and invasive operation that may often be
unnecessary. However, the FAA has a
congressional mandate to make sure that older aircraft are structurally
sound. Non-destructive inspections of
aircraft structures are not sufficient for ensuring that flying aircraft will
not have a structural failure. This
leaves the FAA with only the option of estimating how much stress an aircraft
has undergone per hour of flight and setting life limits in terms of flight
hours.
These estimations are based
on theoretical models whose results vary widely with small changes in their
inputs. One of the key factors in these
models is the amount of stress that an aircraft experiences, on average, per
unit time. Another key factor is how
conservative to be, i.e., how much to err on the side of rebuilding or retiring
a structure early. This project aims to create a device that would record some
indication of how much stress was encountered per unit of time an aircraft was
operating. If the FAA accepted this
data, aircraft that experienced less than average stress could fly longer
without undergoing expensive rebuilding.
On the other hand, some aircraft will experience great stress at times
and may need be rebuilt even before the currently contemplated conservative
models would indicate. Thus, use of
stress-weighted flight time recorders might result in most aircraft being able
to fly for much longer before rebuilding and some aircraft that are prone to
failure as the result of unusual stress being repaired before a catastrophic
failure. Data from aircraft monitored
with stress recorders could also be used to calibrate the models used for
non-equipped aircraft.
Such devices already exist
and are used by the US Navy and the US Forest service to measure the stress on
their aircraft (see the Airframe Cumulative Fatigue Monitoring System
(ACFMS) http://sysei.com/). It
seems to me that SEI does a fine job serving this market, however these systems
cost tens of thousands of dollars per aircraft. I speculate that the reason for these high costs are: (1) they
have to be FAA-certified to be installed in aircraft, (2) FAA-certification
takes years, (3) the number of units sold is small, (4) the number of producers
of these units is small, and (5) the customers are agencies of the US
government, which is used to paying high prices for specialized equipment from
suppliers who can meet onerous government procurement regulations. The high cost of these stress recorders
precludes their wide use on general aviation aircraft such as twin Cessnas.
This project started in the
fall of 2004 as a personal effort. My
motivations are several: (1) I eventually want to see most GA plans outfitted
with flight data and voice recorders, to learn more from accidents and thus
reduce their frequency, and this stress recorder is a good stepping stone for
this, (2) I enjoy building embedded computer based devices, (3) I would like to
explore the possibility of forming a profitable business involving my twin
passions for electronics and flying, and (4) I am interested in owning and
flying twin Cessnas which are older aircraft with stress fatigue issues.
I posted a suggestion for
this project to the Cessna 300/400 forum on the Cessna Forums (http://www.cessna.org/forums/). Mike Busch, a leading Cessna owner and
writer (http://www.savvyaviator.com/bio.html
), suggested that such a device would sell well if it were inexpensive. I have since exchanged email with Mike,
people at the FAA Small Plane Directorate, and with people at SEI (see above).
My interest is more to see
that such a device is built rather than wanting to make money from it. If SEI or some other firm would build what
we need, at this point I would be happy to see them do it and would do what I
could to help them. However, SEI seems
well suited to serve organizations like the US Navy and the cost structure that
requires is, I think, incompatible with the needs of the twin Cessna fleet.
As of early 2005, I have
purchased the components for a prototype and connected some of them
together. The prototype is based on an
ARM-based microcomputer (http://www.embeddedx86.com/epc/ts7200-spec-h.html),
a Trimble Lassen SQ GPS receiver (http://www.trimble.com/lassensq.html),
and an Analog Devices 2-axis accelerometer (http://www.analog.com/en/prod/0%2C2877%2CADXL202%2C00.html). The microcomputer runs Linux. I have it getting GPS data from the Trimble
GPS once per second, two axes of acceleration ten times per second, and two A/D
inputs ten times a second. All of this
is compressed with zlib (http://www.gzip.org/zlib/index.html
) and recorded in a 256MB CF card. The
microcomputer also drives a two line LCD on which I currently display the GPS
time, lat/long, one axes of acceleration, and one channel of the A/D’s
output. My development system is an
Opteron-based Linux system (http://fedora.redhat.com/index.html
) with the ARM cross development tools.
The control program is written in C on top of Linux. I am writing the
control program with the idea that it might run without Linux in the future,
accessing all the I/O devices directly.
I wrote a device driver to measure the pulse widths. The microcomputer accesses the Opteron’s
file system via NFS for ease of development, but all the software can also run
from the microcontroller’s CF card.
I am talking with Strain
Measurement Sensors (http://www.smdsensors.com/)
about how to measure the strain on the critical part of the twin Cessna wing
spars. There are a number of open
questions in this regard, including whether there is access to a part of the
wing for the installation of a strain gauge, what kind of strain gauge to use,
how difficult it will be, whether it will work in the super cold that wings can
experience in northern latitudes in the winter, how reliable the gauge will be,
how much training the installer will need, and how to interpret the results
from the gauge.
Here is an example of the
data from two seconds of operation from the prototype sitting idle. The number after the long/lat is the
altitude (I have changed the altitude values so as not to advertise my home’s
location to everyone on the Internet).
The strain inputs are, for now, just connected to a 3.3V power supply
and ground. I have not yet connected
the prototype to a real strain gauge. I
may need to increase the sample frequency for both the accelerometers and the
strain gauge inputs. The units for the
accelerometers are arbitrary. The
strain numbers are in volts. This data
compresses to be about 21 bytes per second.
It remains to be seen how well typical data compresses. There are certainly improvements I could
make to make the data compress better which might compensate for the expected
decrease in compression efficiency with real data. If I can maintain about 21 bytes per second, I can store about 13,000
hours of flight data in a 1GB flash card.
05:56:10 38:45.9
N 121:47.0 W 370 xg=-14 yg= 7
strain1=3.289 strain2=0.003 xg=-15 yg= 7
strain1=3.290 strain2=0.004 xg=-14 yg= 7
strain1=3.290 strain2=0.003 xg=-15 yg= 8
strain1=3.289 strain2=0.002 xg=-15 yg= 7
strain1=3.290 strain2=0.004 xg=-15 yg= 8
strain1=3.289 strain2=0.003 xg=-14 yg= 7
strain1=3.290 strain2=0.003 xg=-14 yg= 8
strain1=3.289 strain2=0.003 xg=-15 yg= 8
strain1=3.289 strain2=0.003 xg=-15 yg= 7
strain1=3.290 strain2=0.003 xg=-15 yg= 7
strain1=3.289 strain2=0.003 05:56:11 38:45.9 N 121:47.0 W 370 xg=-14 yg= 7
strain1=3.290 strain2=0.003 xg=-15 yg= 7
strain1=3.289 strain2=0.003 xg=-14 yg=
7 strain1=3.288 strain2=0.003 xg=-14 yg= 8
strain1=3.288 strain2=0.003 xg=-15 yg= 8
strain1=3.290 strain2=0.003 xg=-15 yg= 7
strain1=3.289 strain2=0.003 xg=-15 yg= 8
strain1=3.289 strain2=0.003 xg=-15 yg= 8
strain1=3.290 strain2=0.003 xg=-15 yg= 7
strain1=3.290 strain2=0.003 xg=-15 yg= 6
strain1=3.293 strain2=0.003
The reasons that I plan to
measure GPS position and time are that recording these values helps in figuring
out where and when high stress events happened. The Trimble GPS receiver takes almost no power or space and adds
perhaps $70 to the cost of the recorder.
Probably the biggest burden from the GPS is finding a place for the
antenna. The recorder could optionally
take a data stream from an existing GPS navigator in the aircraft. This option would mostly save on
installation cost in the case that taping into an existing data stream is less
expensive than finding a place for another GPS antenna.
Besides the direct value in recording
aircraft stress, fleet operators may also benefit from knowing more about how
pilots are flying their aircraft. For
example, some pilots might be making unnecessarily hard landings, or be pulling
up too hard, or not slowing down enough in turbulence. With a stress recorder, the fleet operators
can get a better handle on this and thus better counsel pilots. This would hopefully lead to less stressed
planes and perhaps safer flights.
My plan to is to build a
prototype, get people interested, hopefully not spending more than $10,000 in
the process (I’m about $2,000 into it so far).
My hope is to then sell about 20 units to the FAA, aircraft companies,
or fleet operators for whom the immediate benefits of the engineering data
would be useful and who could deal with an uncertified unit. Sales of these units would allow me to
proceed to getting an STC. As this
process can take years, I would certainly need to keep my “day job” for a long
time. Another possibility is at some
point to get funding from a foundation or a collection of donors or angel
investors. If others came along who
could do a better job for whatever reason, I anticipate that I would be happy
to pass the baton to them. Yet another
possibility is to sell a version of the recorder as a portable, non-certified
unit, to aerobatics pilots wanting to get a complete picture of their position,
acceleration, and perhaps a voice channel for post-flight analysis. Pilots could compare their flight paths and
the accelerations they experienced with those of top aerobatics pilots doing the
same maneuvers, for example. This might
not be a big market, but perhaps big enough to get some experience and capital.
I have spoken with the FAA
about this project. They have been
quite helpful and mildly encouraging. I
think they have a lot of people who contact them and who get excited about
building something, but don’t follow through.
They said a number of useful things, here are three of the most salient:
In 1980 I was a passenger in
my brother’s Cessna 172 when we crashed and he was killed during an instrument
approach in bad weather. As a result of
that experience and my long interest in building electronics for social
purposes, I am motivated to build electronics to make general aviation safer. I believe that GPS navigators, TAWS (Terrain
Awareness Warning System), fuel totalizers, traffic warning systems, and other
innovations, along with better training, will eventually lead to much lower
accident rates. Although progress in
these areas is slow to those of us in the technology business, it is
happening. One area where I see no
progress is in getting more information for the approximately 300 fatal general
aviation accidents that happen each year in the United States.
The NTSB investigates every
accident and generates a report for each one.
However, reading these reports is of marginal value for pilots. They have a lot of detail about what the
accident investigators found, but are usually completely lacking in the
“why”. At first, reading these reports,
one can think about many of these accidents “what a dumb thing that pilot
did”. But after a while, one realizes
that many of the accident pilots were not unusually bad pilots, they were
perhaps like most of us and unfortunately often made one or a series of bad
decisions. I believe that if we had
more information, information that can be recorded cheaply with modern
electronics, we could learn more from accidents. The result, hopefully, would be fewer accidents as the rest of
could, with the help of NTSB interpretation, learn more of the underlying
causes of fatal small-plane accidents.
To get aircraft owners to
install flight recorders, several things must be true:
I believe that with modern
electronics and a modest set of parameters to measure, that the cost could be
made fairly low. Here are the things
that I propose a low cost flight recorder measure:
There are many other
parameters that would be interesting to accident investigators, but these are
the ones that I believe are (1) useful, and (2) cheap to measure and
record. As to survivability, I suggest
that such a recorder have enough insulation to survive a modest fire and
submersion in water, but that it not go to the lengths that the airline
recorders go to if that makes the unit too expensive to build or install. Sometimes the tough recorders used in
airlines do not survive and certainly the low cost recorders I am suggesting
would not either. They do not have to
survive all accidents, just most of them.
The flight recorders used on commercial airlines cost, I believe, about
$130,000. Here is more information on
these: http://www.flightrecorder.com/pages/1/index.htm.
Another possibility is to
integrate the flight recorder functions into an existing navigator, TAWS, MFD,
or some other unit in the aircraft.
That would make it less expensive, perhaps just the cost of some
non-volatile memory wrapped in a ceramic blanket and mounted in the tail for
greater survivability. It could connect
to the existing avionics via a USB cable.
I both started building
embedded computers and flying airplanes in 1976. However, I didn’t get my pilot’s license until 1998. I have about 1000 hours of total flight time,
instrument and multi engine ratings, and a commercial license. For the last three years I have owned and
flown a Cirrus SR22. However, I sold my
four-seat SR22 and I am looking for a twin Cessna. I have spent many years in the computer industry and in academia
working on computer systems. I have a
BA and Meng from Cornell in Computer Science, a Ph.D. in Computer Science and
Engineering from the University of Washington (in Seattle). I was a postdoctoral associate at MIT ten
years ago. I have since worked at a
chip design startup and a company that designs and manufactures
microprocessors.
If you would like to comment
on this effort or these web pages, please contact me at flightrecorder@bedichek.org.
Robert Bedichek
January 5, 2005