l"Suppose you have line maintenance done on a defect—unscheduled maintenance—and you need to fix it and get the aircraft back into the air,” Kanishka Agiwal, associate partner in IBM’s Watson Travel & Transportation unit, says. The line technician fixes the problem and makes some quick notes, using his natural language in a log. But what happens the next time the same problem occurs, perhaps with the same part? 

Agiwal notes that by using Watson and Natural Language Processing, IBM can interpret all those natural language notes, called unstructured data, and help the next engineer fix the problem faster or even anticipate the defect, turning unscheduled maintenance into much less costly scheduled maintenance. For example, an inflight entertainment system in a particular seat may malfunction regularly because it is being hit by a trolley. Interpreting unstructured repair notes can discover this cause and find the most effective prevention or repair strategy. “We are democratizing knowledge,” Agiwal says. “We can make a rookie engineer as good as a veteran one.” The technique works on a variety of components, from hydraulic systems to flight controls and flaps.

IBM has already used the technique with several airlines and helped to reduce line maintenance time by 15-20 min., increase the availability of the right parts by anticipating needs by line station, and reduce those frustrating and expensive “No Fault Found” (NFF) efforts by 3-5%.

The IBM system can also interface with the airline’s enterprise research planning (ERP) system to get the supply chain moving faster. It works with unstructured notes by pilots, mechanics, engineers and other staff. And it can combine unstructured data such as words, pictures and video with the massive structured data generated by sensors and transmitted by ACARS (Aircraft Communications Addressing and Reporting System) to yield even more significant results. Watson understands natural phrases in context, in many languages. “I don’t think anyone else comes close to what we can do with unstructured data,” Agiwal says.

The natural language tools could be added to predictive maintenance tools IBM is helping Airbus and EasyJet develop. IBM consultants will discuss achievements and future potential at a Watson conference in San Francisco on Nov. 9-10.