This is the second in our three-part series about how big data and analytics are transforming the face of operations across industries. In the first part, we focused on how analytics is driving operational excellence. This part of the series provides our point of view on the use of analytics in the aviation industry.
Airlines incur huge margin erosion and face customer dissatisfaction from flight delays and cancellations—in the US alone, 650 million passengers fly every year. The major reason is aircraft on ground (AOG), a situation in which a plane is down for want of a critical spare part, which could be due to a snag that occurred during the previous flight. 40% of all aircraft delay is caused by technical issues (in India, it's close to 50%) and the associated spare-part shortage, which results in an $8 billion loss per year in an industry already plagued by huge operating costs.
To address this issue, airlines are combining digital technology and analytics to improve the accuracy of their forecasting and availability of the right spare parts to reduce delays and improve margins.
Aircraft are made of complex systems with many computers and sensors that generate alert codes, which get transmitted to the ground continuously or intermittently, as decided by the airline. These codes, also known as dispatch/non-dispatch messages, contain information on issues that can develop at any time during a flight. Normally, ground engineers receive and decode the message, try to identify the reason for the snag, and decide the maintenance action to rectify it. This needs to be done in just 40 minutes and relies on replacement parts being available. More often than not, they aren't.
We have recently worked with a few airlines to improve how they plan for parts proactively. With data available in abundance from hundreds of daily flights, this is a typical big data problem. They have structured data, like dispatch/non-dispatch messages, unstructured data, such as maintenance logs, and component-change history, which combines airborne and ground operations data.
With the right technology stack in place, we established the relationship between the dispatch messages, the associated snags, and the correct remedial action to be taken.We created an early warning system for non-dispatch event types using historical data. Reliability analysis using historical data provided the failure rate and failure behavior. A two-step hierarchical Bayesian analysis with random forest modeling was deployed using Hadoop and R, which first forecasts the probability of there being a non-dispatch message during the flight and then forecasts the probability of a spare being available to rectify the snag.
With rigorous testing and validation, the solution provides 75% accuracy for forecasting the replacement of a spare part, reducing delays and cancellations significantly. Using digital and analytics solutions in this way can reduce downtime for airlines, greatly improve performance, drive down costs, and bolster passenger satisfaction through more frequent on-time arrivals and departures.