Increasingly, You're Being Watched -- Why That's Good
The NTSB's recent release of a final report regarding the fatal November 13, 2011, crash of a Cirrus SR-22T shed light on a flight noteworthy for its aerobatics; it also invites speculation regarding the potentially untapped value of recorded flight data in general aviation.
Both Avidyne and Garmin produce avionics (the Entegra and G1000, for example) capable of recording tens of different flight parameters that the entire aviation community -- let alone individual pilots -- may not be using to full advantage. In the case of the Cirrus crash, data stored in a remote (tail mounted) data storage unit supplied sufficient information to not only reconstruct the aircraft's activities immediately prior to the crash, but also for several days before. One man in particular, outside of the NTSB, made unique use of that data. And hopes to play a part in the charge for the release of more data in other crashes. His name is Rick Beach and what he hopes for would help us all.
Particular to the November 13, crash, the aircraft, a rental, recorded information from its Avidyne and other systems to a recoverable data module (RDM). Beach told us the folks who make the RDM for Cirrus are a company called Heads Up Technologies. The company makes a "single chip processor capable of addressing large flash memory arrays" that is contained in an "armored housing" and mounted somewhere in the aircraft that isn't likely to hit the ground first or be consumed by a post-crash fire. Usually, that means the vertical tail. On a normal day, the data can be downloaded via a standard USB cable. In the case of a crash, the NTSB can of course access data in several other ways that don't involve the standard USB plug. In this case, they recovered very telling information.
Recovered data led the NTSB to determine that the crash was the result of a roll attempted at low altitude. The NTSB used the information to describe the moments leading up to the crash and also its recent flight history, which involved aerobatic maneuvers while the aircraft was in the care of the same pilot. The agency made some of the accident aircraft's recorded data publicly available, and that's where Beach comes in. Beach mined it. He took the data points and created a virtual presentation (a video created in a flight simulation program) of the event -- something he's done for roughly 15 other accidents. He then took that video and posted it to Youtube as well as Cirrus-related forums. But it's the "why" that makes all the difference.
Life Beyond The Data Points
Beach works as an unpaid volunteer contributor to the Cirrus Owners and Pilots Association (COPA). He distributes the data in the form of video presentations for educational purposes. Beach believes that the graphic display of events delivers to pilots a more powerful message than the "pilot failed to maintain altitude during maneuvering" more typical of an NTSB-published analysis. "I showed the animation of a dual fatality 10-turn spin into the ground. To see the plane slow and wallow in mush condition ... then you hear the power come up and the plane spins. The engine noise never decreases. It takes 25 seconds to play it." Beach says that the visual effect is different than a factual report in important ways. "When you say '25 seconds' that's just math. In my presentations, I'm silent during those 25 seconds. It delivers the reality of the event. These guys had 25 seconds of really bad situation and they didn't pull the chute until it was too late. It brings a lump to my throat." And there-in lies the value. Beach believes the real impact of transposing the data into a virtual experience delivers something that leaves an impression. And it's that impression that can change how pilots think and behave in the cockpit. ... And he has some numbers that may support that.
Cirrus recognized the importance of data mining "about 2007," Beach says. Since the aircraft could survive after a chute pull Beach says Cirrus wanted a standardized way to acquire the data in all accidents. As a result of that thinking they introduced the crash hardened data recovery module into their airframes. Then they worked to apply what they found.
According to Beach, "during the second half of 2011, there were 11 fatal Cirrus crashes and only one CAPS (Cirrus' full plane parachute) save." For the same period in 2012, "there were six fatal accidents for four CAPS saves." Beach attributes that improvement to awareness and training derived from data mining. From graphic presentations (like these provided by COPA), that conclusion may not be immediately clear. Testimony is another matter.
Of the four 2012 saves, the surviving pilots each credited their training and the Cirrus Owners and Pilots Association community for preparing them to take action when action needed to be taken. At least one pilot specifically noted a video of one of Beach's live presentations. According to that pilot, "I sat and watched it over lunch and without it, I doubt that I would have pulled the chute. After watching all those outcomes, you can put yourself in those situations, consider your criteria, get a realistic image of do or don't." The training replaces real-time contemplation shown to get pilots killed, says Beach. "When stressors are introduced, that's a bad time to start thinking about pulling the chute." But Beach believes data mining could easily go well beyond CAPS deployments.
"The future," says Beach, "is to gain access to even more data, so that there can be more investigation of pilot behavior." In an ideal world, Beach sees data mining driving intervention-style training adapted to correct -- not punish --specific pilot behaviors. This approach, he feels, "would address the common concern: that most accidents are caused by pilot related actions." Beach believes more data could shed more light on the trail prior to the accident chain, be it poor training, bad habits, or behavioral dispositions. In a broad sense, says Beach, "the data gives us the ability to better understand what pilots do wrong."
Beach sees a world where data mining helps flight departments improve pilot training and performance as part of a safety management system. He sees the possibility of using the data to identify problematic patterns and develop intervention-style training to address those issues. Beach believes he's already seen this approach applied with some success regarding Cirrus owners and CAPS use. And it can certainly be used in initial training by flight instructors to address the details of a flight with a student and apply corrective instruction. The prize, according to Beach, goes further.
With enough data collected from the routine flights of pilots who eventually crashed, it could be possible to identify pre-existing problem areas that ultimately contributed to the crash scenarios. As Beach puts it, "was the crash a unique event, or was there training, or something else, that contributed to the crash well before the crash sequence was initiated?" Data mining could help answer those questions. It could help identify weak spots in training, or even in the routine flying habits of individuals, that ultimately contributed to crashes. That, in turn, could lead to improved training for all pilots, or safety interventions for specific pilots whose own flight data matched problem patterns shown by that of crash pilots. With enough available data and common software, pilots could identify their own potential problem areas and seek correction. But there are barriers.
In his quest for more data, Beach says he sometimes feels hampered by the amount of information the NTSB makes available for release. "If we get the information form the NTSB, they typically have put a window on what timeframe they analyze," he says. Beach's concern is for the extra information that may also be revealing. "Sometimes the extra information is there, but we don't have access to it." AVweb reached out to the NTSB to ask what determining factors are involved regarding the release of information. Here's what we found:
Joe Kolly, director of research and engineering for the NTSB, told us "the NTSB recovers data from any device that may contain data of interest to the investigation." Kolly said that the agency assesses "the potential value to the investigation and the effort required to retrieve the data (for instance if it is damaged)" as the basis for its selection. "Any data that we use from these devices is written in a laboratory report, and presented as a series of plots." It is then reviewed finalized and placed in the public docket for the accident. Any data downloaded and retained but not made public may be requested through the Freedom Of Information Act, Kolly said. He added, "Release of any such data is subject to the policies and law governing these types of requests."
Mining Your Own Data
We reached out to both Avidyne and Garmin to ask about their current involvement with respect to data mining -- specifically for improved flight safety. Avidyne had to cancel and told us to talk to Rick Beach. Avidyne, he said, once performed an internal study of every Cirrus accident involving an autopilot, "so they have an intimate understanding of data mining for safety." As for individual pilots, data can be collected downloaded and stored, or not, says Beach. Websites like CirrusReports.com allow pilots to share data online to create a database or seek help for problems. At Garmin, Tim Larberg (field service engineer, aviation product support) and Joey Ferreyra (aviation marketing manager) told us that, for the most part, users are working independently with data captured by their G1000 units. Generally, they said captured data is used to help identify and diagnose things like intermittent engine problems and maintenance issues.
Garmin developed recording capability for the G1000 about three years ago. It requires the user to plug an SD card into the top slot and the unit will record some 30 parameters each second of every flight. A one gigabyte SD card should handle 1000 flight hours, Larberg said. And each departure gets a time stamp with the departure airport as part of the file. As with Avidyne, downloading data for off-aircraft storage is up to the user. Garmin offers free software on its public website that allows users to display data on google earth with a bar graph for each second. There is no little airplane flying around as in Beach's recreations, but each particular bar gives users all the parameters for that point in time.
Using data to sort out maintenance concerns may be the present. But, as Beach shows, data mining to improve flight safety through pilot training and safety intervention is one possible future. Toward that end, if you're moved by Beach's arguments that such improvements could be facilitated through broader access to data from crashed aircraft, the NTSB's Kolly offers a course of action. "Comments and suggestions are always welcomed at the NTSB. Perhaps the simplest means is to visit our website, and submit comments there (http://www.ntsb.gov/about/contact.html). Or follow the instructions on that page to send a letter directly to our leadership."