Loading

How I Learned To Stop Worrying And Love The Singularity Using Collective Data to Drive Safety Improvement

by James Williams, FAA Safety Briefing

In technological terms, we define the singularity as a hypothetical moment in time when technology, particularly information processing and artificial intelligence, becomes so advanced that humanity undergoes a dramatic and irreversible change. In science fiction, this is usually a moment of terror when our new robot overlords rise to power. In reality, however, this moment is unlikely to be so clearly defined or as ominous. In fact, we’ve already seen something of a singularity occur in aviation.

In my youth (which wasn’t that long ago), airline accidents were not common but still occurred on average a few times a year. After each accident, the government would remind us just how much safer air travel was than it had been even a generation before. These statements were unequivocally true. The situation was so much better that looking forward wasn’t easy. The metaphorical low-hanging fruit was long gone, and the law of diminishing returns was clearly setting in. The current system of regulation and safety was approaching its theoretical limit.

Synergy is Not Just a Buzzword

Synergy is a buzzword when used in place of an actual strategy. Happily, though, in this case, synergy was key to advancing safety. That synergy took the form of an FAA/industry collaboration called the Commercial Aviation Safety Team (CAST).

CAST uses a data-driven strategy to reduce the fatality risk in commercial air travel. This cooperation has led to historic reductions in the commercial aviation fatality risk.

The CAST model was the driving force behind an 83-percent reduction in the risk of fatalities between 1998 and 2008. But once you drive accident rates so low, how do you proceed?

The FAA’s response was to move to the Aviation Safety Information Analysis and Sharing (ASIAS) system. ASIAS takes data from voluntary programs like the Aviation Safety Action Program (ASAP), Flight Operational Quality Assurance (FOQA), and the Air Traffic Safety Action Program (ATSAP) and combines it into one database. Through ASIAS, the FAA and industry can look for risks in the system and work to mitigate those risks before they become accidents. This means there is even more data to better monitor risks and determine interventions.

ASIAS has been a great system for all involved and is continuing to evolve. Along those lines, the FAA’s Center of Excellence for General Aviation, PEGASAS (Partnership to Enhance General Aviation Safety, Accessibility, and Sustainability), has two active projects focused on new ASIAS communities.

Cracking the Helicopter Conundrum

It’s no secret that improving the safety of helicopter operations has been a goal of the FAA for decades. But the challenges that face the FAA and industry are daunting. “There are all kinds of unique operations from air tour and corporate flights to heavy lift, pipeline/powerline inspection, law enforcement, training, and aerial application just to name a few,” explained Cliff Johnson, an engineer from the FAA’s Technical Center in Atlantic City, New Jersey. “They are all unique and face their own challenges during normal operations. The other aspect has been the relative lack of flight data monitoring being used across the industry,” Johnson continued.

It’s no secret that improving the safety of helicopter operations has been a goal for the FAA for decades.

Johnson is the FAA Technical Contact for the PEGASAS project on Helicopter Flight Data Monitoring (HFDM). “Since we’ve started the project, we’ve seen more operators embrace HFDM technology, but by and large the technology is still in its infancy for this community, except for certain mission segments like oil and gas or offshore,” Johnson said. “That is starting to change with the new 2018 Helicopter Air Ambulance (HAA) rule and as more operators gain knowledge of the systems and their capabilities. I like to think that our project has contributed to helping spread the word about the benefits of HFDM technologies and how they can be used during daily operations.

“We began this effort back in 2014 and we’re currently continuing to develop some of the metrics, events, and exceedances to analyze helicopter data,” Johnson explained. “These capabilities will form the backbone for incorporating and analyzing helicopter flight data in ASIAS, which is expected to take hold within the next several years, as we transition research capabilities and prototype data analysis tools into operational modules.”

That incorporation into ASIAS is critical to safety improvement. To develop effective safety enhancements, we have to have data to define the problems and measure our success in dealing with them. Much of the work to get helicopter data into ASIAS revolves around the variety of helicopter operations Johnson mentioned earlier. But that problem is only the half of it. The other half is getting more operators to install HFDM devices.

Thanks to modern avionics, FDM in rotorcraft and light GA airplanes is now more economical.

The United States Helicopter Safety Team (USHST) is on board and has issued a Safety Enhancement to support HFDM equipage. “One of the USHST helicopter safety enhancements [H-SE 82] promotes installation and use of HFDM devices for the purposes of detection and monitoring of aircraft and engine limitations that were exceeded, collecting and preserving more data relevant to accident investigation, and detecting and correcting procedural noncompliance,” Johnson explained. “This offers operators the opportunity to incorporate HFDM into their own operations, which will increase the safety and efficiency of flight operations and hopefully reduce the helicopter fatal accident rate.”

Pooling data could be a powerful tool for the rotorcraft community because there are many smaller operators that might not have a large enough fleet to detect systemic issues on their own. But to do that requires a good system to manage it, and a large supply of data to fill that system. The PEGASAS HFDM project hopes to help on both fronts.

Seeing the Forest for the Trees

We’ve discussed previously in this magazine about how modern avionics have made flight data monitoring (FDM) much more accessible to GA. But the question is how do GA pilots access and use that information? Unlike in the commercial world where you have a structured system like FOQA that can be easily used to tap the data from operators, the GA community has more limited options, despite its much greater footprint on NAS operations.

To provide a solution to that problem, the FAA partnered with academia and industry to create a portal that could collect data from the wide variety of GA operations. The end result was the National General Aviation Flight Information Database (NGAFID). The NGAFID enables pilots to upload data from their avionics or from a specially designed smart phone/tablet app. This is an easy, no cost way not only to examine your own data, but also to share it with the database. As with FOQA, the individual data is de-identified and cannot be used by the FAA for enforcement.

Images from the NGAFID that allow users to view the data from their flights.

The General Aviation Joint Steering Committee (GAJSC) is working to spread the word on the benefits of NGAFID and has signed up 11 universities and 70 corporate flight departments in addition to individual GA pilots. In total, more than 810,000 hours of flight data have been collected in the light GA community alone.

“Flight schools are quickly realizing the benefits,” Operations Research Analyst Corey Stephens said. “The more all of us work together, the better.” Stephens hopes to see similar safety improvements to the ones seen following the establishment of CAST.

The second problem is very similar to the one faced by the rotorcraft community: with such a diversity of operations, how can the data be organized for effective use? We need a meaningful way of analyzing all the data we acquire. The 11 universities involved are doing great work incorporating that data into their SMS (Safety Management System) programs, but how do we export those lessons to the wider community? That’s where the PEGASAS Safety Analysis for General Aviation project comes in.

“This project develops prototypes that, once fully developed and implemented, could provide all pilots with tools that were previously only available to pilots of GA aircraft with advanced avionics and flight data recorders,” explained Professor Karen Marais of Purdue University. Marais is the Site Director at Purdue University and the PEGASAS Technical Contact for the project.

“The introduction of smartphones and tablets to the cockpit for EFB [electronic flight bag] purposes, as well as the addition of AHRS [attitude and reference heading system] devices, can give us insight on how pilots fly in the absence of data of higher resolution,” Marais continued.

The introduction of smartphones and tablets to the cockpit for EFB purposes, as well as the addition of AHRS devices, can give us insight on how pilots fly.

“The use of machine learning adds to the data by training algorithms on high-resolution data, and then applying the same algorithms to data that tablets and smartphones would record. As a result, not only can we identify areas that are deficient in aggregate and inform decision-making, pilots can also learn from their flights and improve their safety.”

Analyzing data can sometimes lead to even more data. Professor Marais explains how this happened with the project. “The ability to retroactively ‘recover’ data that was not originally present in the dataset was a nice outcome from the GA Safety Analysis project. For example, while most GA flight data recorders do not record flap position, feeding the available flight data into aerodynamic and propulsion models and comparing what we see versus what we were expecting to see, allows us to infer what flap settings the pilot was using during the flight. This work was accomplished by our team members at Georgia Tech.”

This kind of post hoc “improvement” of datasets give them more power by letting us see factors not in the original data. For instance, is there an increase in takeoff accidents or incidents in aircraft X at flap setting Y? It’s probably not going to be so simple in real life, but the point is that such a trend would be invisible without the “recovered” data.

We can see that PEGASAS is working to not only improve the high-resolution data that we do get, but also to better analyze the low-resolution data. The latter point is important because low-resolution solutions are far easier and cheaper to install in most aircraft. Removing the barriers to entry is critical to expanding the data pool.

“Due to flight data sharing sensitivity and ASIAS protocol, at this time we have only used a relatively small set of data from Purdue University flights, as well as a subset of flight data provided by the University of North Dakota, for our research purposes,” Marais said. “Based on this limited set of de-identified data, we have observed that students in flight programs tend to develop good habits through their flights. Any abnormalities present in the flight data tend to be small and easily corrected.”

The takeaway here is that the more data we can feed into projects like this, the more we can improve safety in GA.

The Singularity

We need both data and an intelligent analysis but these concepts are clearly interdependent. In both the GA and rotorcraft realm, we are late to the FDM game.

We need improvements in both the quantity and quality of the data we collect. Likewise, we need to improve the analysis of that data in terms of quality. PEGASAS is hard at work in both communities.

The end goal is a singularity of sorts. By increasing the amount of high-quality data we have, and improving the tools we use to analyze that data, we can create our own singularity. We can create that irreversible change in a positive way.

More data means more safety issues detected. More safety issues detected means more potential interventions. That cascade of events should lead to fewer accidents and fewer lives lost. I think that’s a far nicer singularity to look forward to than the one envisioned by science fiction.

James Williams is FAA Safety Briefing’s associate editor and photo editor. He is also a pilot and ground instructor.

This article was originally published in the May/June 2018 issue of FAA Safety Briefing magazine.

Created By
FAA Safety Team
Appreciate

Report Abuse

If you feel that this video content violates the Adobe Terms of Use, you may report this content by filling out this quick form.

To report a Copyright Violation, please follow Section 17 in the Terms of Use.