light delays and cancellations caused by Reducing Unscheduled Maintenance Delays and Cancellations using AI/ML 1 2 3 4 5 Reviewed for technical accuracy March 10, 2021 © 2021, Amazon Web Services, Inc. or its affiliates. All rights reserved. 7 6 AWS Reference Architecture unscheduled maintenance issues cost airlines $120K —$300K per aircraft per year; potentially translating to $60M — $150M annually for an airline with 500 aircrafts. Implementing predictive maintenance solutions that leverage the aircraft log, sensor, and maintenance data can reduce this cost by up to 25%. 1 2 3 4 5 6 7 The ACARS system collects fault logs and pilot reports in real-time. QAR data is collected and uploaded into AWS based on aircraft capability. The frequency of data collection varies based on the connectivity package installed on the aircraft. The ACARS system can send selected signals on a limited basis. Maintenance logs, part removals, and part repair history from the maintenance & engineering systems are collected periodically based on the capability of the system. The flight operations system provides the scheduled and actual flight information. This is critical for correlating delays and cancellations with component failures. Identifying information must be deleted from the QAR data before it is used. Amazon SageMaker trains the model to correlate delays and cancellations to fault logs, maintenance logs, part removals and QAR data. Feature engineering identifies the most significant and predictable chapters and components. Models are tuned peri
