Not long ago, on a Sunday night like any other, I found myself waiting in a terminal at the Charlotte airport for a connecting flight. As most 21st century travelers have come to expect, the boarding process didn’t begin on time-but as most of us dreaded-our momentarily postponed boarding queue turned into two hours of maintenance on an unspecified piece of equipment. We finally took off as midnight rolled us over into the next day, and arrived at our final destination safe, but exhausted.
At face value, maintenance seems to be the most forgivable reason for a delay. FAA regulations mandate that even the simplest repairs must be made before a plane leaves the ground: if a latch on an overhead bin were to remain broken for the next flight, the airline could potentially be responsible for a serious injury to a flyer from an airborne piece of luggage. However, as forgiving as flyers may choose to be when faced with maintenance delays, the fact remains that delays make no one’s day. Unfortunately, maintenance can prove quite an extensive endeavor: in fact, data from MasFlight indicates that about one-third of flight delays and cancellations are maintenance-related .
But what specifically happens during maintenance to prolong the process so much? According to one helicopter pilot and former flight quality assurance specialist, a “maintenance delay” can look like this:
1. A problem is identified by a pilot or line maintenance crew doing a pre- or post- flight inspection, or was identified during the flight.
2. The maintenance procedure manual is referenced for a solution to the problem.
3. The maintenance technician collects all the tools and equipment needed to fix the problem, which are listed in the procedure. Lots of special tools, diagnostic equipment, and parts are needed for aircraft maintenance. They are often expensive, and sometimes hard to find.
4. The technician fixes the problem.
5. The technician accounts for all tools.
6. The technician records the completed procedure in the aircraft’s maintenance logbook.
7. The technician locates the quality assurance specialist.
8. The quality assurance (QA) specialist checks the technician’s work. The FAA requires this check.
9. When certain components are replaced, they require checks to be completed by the maintenance technician or pilot to ensure the new component is installed and working properly. These checks are discerned and completed as needed.
10. After these checks are completed and the aircraft is deemed ready, the QA specialist indicates completion in the aircraft logbook.
One of the earliest steps in this process may appear the most unassuming. Checking a procedure manual, to an average passenger, ostensibly seems like it would take the least time and effort. However, locating a specific solution to repair a given part can prove an arduous process. Before aircraft maintenance manuals were digitized in 2000, printed binders were extremely difficult to sift through. A telling example involves the Boeing 777; according to the Boeing website, the printed version of the primary Aircraft Maintenance Manual for the 777 “takes up 24 binders and requires 10 feet of shelf space.” Since digitizing all manuals, this process has been expedited to an extent, but still remains complicated.
“Last year alone, Boeing distributed enough maintenance documents to create a stack of paper more than 24 miles (38 km) high and a stack of microfilm cartridges more than 14 miles (22 km) high.”
Furthermore, as planes in an airline’s fleet age, the likelihood and urgency of repairs increases accordingly. Pilots know better than anyone that aircraft parts must be replaced rather predictably after a certain number of flight hours or by a certain date. But this does not account for parts expiring or wearing out before their documented date, which occurs sporadically. The issue of maintenance delays thus exists as a dilemma of both inconvenient categorization and an incapacity to predict the future. In each case, what’s a maintenance engineer to do when faced with an unplanned part replacement or a complicated repair?
Enter artificial intelligence, a resident expert at quickly identifying problems and directing users to a viable solution at a speed far exceeding that of the most experienced human expert. With predictive analytics, the ten or more steps within aircraft maintenance could be reduced down to the repair itself and corresponding QA check. Delays could be mitigated altogether, as fleets could perform maintenance on predicted failures while the aircraft is off the clock. In fact, an endeavor to structure a seasoned aerospace engineer from an artificially intelligent learning environment could suspend a large portion of existing delays and cancellations, if applied effectively. Rather than acting alone, predictive maintenance could implement sensors to accurately track performance statistics per part, in tandem with existing part schedules and catalogues, as well as identifying problems far in advance.
AI could perform the following systematic maintenance augmentations:
- Ask questions that would otherwise would have involved Aviation Technical Service (ATS) reps
- Provide context to ATS reps automatically when they join the conversation
- Track path to resolution
- Track complex presentation of problems; for example, fault codes that show up together
- Add an interface to knowledge base (aircraft maintenance manuals)
- Encourage collaboration by providing a seamless interface to exchange data
Leading AI companies today are deploying such systems. For example, SparkCognition has partnered with Honeywell to develop an IBM Watson-powered “advisory” application for aerospace maintenance. The goal of this partnership is to deliver an artificially intelligent approach to flight maintenance in order to optimize workflow and deliver relevant documentation for faster repair turnaround. Such AI-powered systems can monitor the health of aircraft with health prognostics and deliver predictive maintenance insights while providing an in-context advisory that can go as far as to align job assignments to match technician experience and expertise. The results are optimized schedules, minimized maintenance costs, maximized safety, and most importantly, avoidance of flight delays.
SparkCognition has partnered with Honeywell to develop an IBM Watson-Powered “advisory” application for aerospace maintenance. The goal of this partnership is to deliver an artificially intelligent approach to flight maintenance in order to optimize workflow and deliver relevant documentation for faster repair turnaround.
By closing the gap between a digital repository of maintenance knowledge and a human engineer executing relevant repairs, predictive maintenance with AI advisory can transform the current shortcomings of the aerospace industry and get passengers home sooner.
This author will confess that thinking about the gleaming, futuristic implications of AI-driven engineers on airline maintenance is certainly most hoped-for while on a flight delayed by maintenance. But if underlying (and oft unspoken) need truly drives the best innovation, then predictive maintenance will be the chance to transform an industry for the better.
by Caroline Lee
Last modified: November 22, 2017