One of the main challenges in aviation is to reduce costs and delays, while maintaining or improving current safety levels. A large percent of these costly delays are a result of unplanned maintenance such as when an aircraft has an abnormal behavior on the previous flight, creating an operational disruption and even requiring an aircraft change. Improving reliability and predicting failure are key aspects for reducing maintenance costs. Considering this, this is a subject undergoing intense study and large companies in the sector such as Airbus or Boeing are investigating it.

Currently, an aircraft inspection happens as needed by periodic maintenance requirements or when abnormal events occur (component failure, damages…). In the first scenario, routine aircraft maintenance can range from daily examinations to tests that occur every five years (corresponding to checks A to D REF). However, can these routine examinations be substituted by a more evidence-based approach? Or, could component failures be prevented beforehand?

To answer these questions, the term “predictive maintenance” must be introduced contrary to the preventive approach that is more frequently used. It requires knowing the correct information at the suitable moment to perform the necessary maintenance, which then requires a combination of domain knowledge and data science skills. In general, this concept consists of predicting when failure will occur and preventing it from happening by scheduling maintenance and carrying out a check. The direct costs, skill level and experience for understanding and controlling the data are higher at predictive rather than at preventive maintenance, but the flip side advantage is that the total time and costs are exponentially reduced. This is the case of the five-year agreement signed between Airbus and easyJet to provide this type of maintenance to the entire fleet (almost 300 A320 family aircraft) starting summer 2019, along with the Boeing-Japan Airlines contract for their 787 fleet, and the ConditionAnalytics and DATCOM products developed by Lufthansa Technik for predictive maintenance.

Big data once again plays a crucial role in aviation. But what are the necessary data and steps for performing predictive maintenance? Two data types are needed: reference data and operating data. Reference data show the normal behavior of the system and operating data show the aircraft’s ideal situation, transmitted either directly in real-time flight or on-ground. Predictive maintenance analysis provides decision support tools to help engineers identify potential failures that may require maintenance. Along this process, data sharing is a key part to refine the predictions and further improve its accuracy. Accordingly, the data availability and the improvement on its recording are key challenges to make the most out of this innovative technique. If you are interested in this topic, and the opportunities, check out the related post: Data Exchange Programmes.

Predictive aviation uses a software program that uses sensors and Flight Data Recorder (FDR) information to show if a failure may occur. The in-flight data is downloaded from the aircraft’s Flight Data Recorder to computer software where irregularities are identified. If any are detected, information is sent to schedule a check. Consequently, planes are less likely to stop working, have delays or incur cost.

Considering this, improvements are being made. Productivity is improving and predictive maintenance is the future of aircraft MRO (Maintenance, Repair and Operation). However, success will depend on achieving three goals: obtaining the right aircraft data, addressing the problem appropriately and properly evaluating the results.