Flight data analysis: why bet on big data solution? - Cassiopée Safran

Global air traffic is booming, your fleets are expanding, and your airplanes are generating more and more data: your systems have to change as well!
But how do you choose the right flight data management software? Make yourself comfortable while we explain the process in four easy steps. 

01 Automation and speed: the dynamic duo

Your first concern, and far from the least, is how fast you can decrypt your data, whether you want to analyze the huge streams of daily flight data, improve flight safety, reduce maintenance costs and fuel consumption, or for any other operational requirement.

What you need is software that’s capable of automatically decoding all data generated by your airplanes. By sparing you the job of manually selecting the files to be analyzed, automation saves a significant amount of your precious time. However, you have to make sure that this advantage is not offset by decoding that’s too slow. But remember to compare what’s comparable: don’t focus on the decoding speed of a single file, but rather on all of your data. Today’s decoding systems adapt to the performance of your own computer system. By using all of your machine’s processors simultaneously, they concurrently execute several instructions and maximize decoding speed. The most powerful tools may even offer solutions that adapt to the intensity of your computation requirement at a given moment. This capability takes us into the heart of big data.

02 Integrate all data formats

Do you deploy a mixed fleet? Do your new aircraft have data formats different from QAR in the ARINC 717 protocol? Don’t skip the analysis of this data! If you’re used to QAR (Quick Access Recorder) or even DAR (Digital ACMS Recorder) formats for your FOQA (Flight Operation Quality Assurance) programs, you undoubtedly already use conversion tools. And if your fleet includes the Boeing 787, you also decode ARINC 767 for your QAR files (this can also be called Continuous Parameter Logging). On the other hand, perhaps you’re not already using SAR (Smart Access Recorder) files generated by the Airbus A320, A350 and A380, as well as the ATR and 737NG. This format, which is not really all that new, is increasingly used for predictive maintenance tools, because of its “intelligence”. SAR files allow you to start a recording based on triggers. You don’t need to record the whole flight; you can now target a specific period. By configuring the recording buffer, you can even highlight the zone to be monitored, and get information before, during and after the event. You wouldn’t want to miss out on this new source of information, would you? As you’ve probably understood by now, you should choose a decoding tool that also includes decoders of ARINC 717 and ARINC 767 files, as well as SAR.

03 HDF5 & XML to facilitate your data analysis

Being able to analyze events by flight and by airplane is no longer enough. Now you naturally want to understand your overall fleet performance, over a given period. The upshot is that algorithms are also changing, becoming increasingly complex and now supported by tools well known to data scientists, such as MATLAB, PYTON and R. So your output formats have to be compatible with these tools. The CSV format, too slow and unwieldy, which was extensively used for many years, is now reaching its limits. The solution is HDF5 and XML files used in unison, the former to access all flight data (altitude, speed, etc.), the latter for its meta-data, namely the flight “ID” (flight number, departure and destination airports, flight length, etc.). Compatible with the most commonly used and advanced mathematical software, HDF5 and XML formats are also far more effective to process. An HDF5 file, for instance, can be read 25 times faster and is 1/85 the size of a CSV file! In short, input data formats are just as important as output data formats.


Decoding, sorting and contextualizing data doesn’t do much good if you can’t handle the data. But, when you have to validate a new failure detection algorithm, you know that you’ll have to test it on replayed historic flight data (spanning a few months to a few years) for your entire fleet.  You therefore need a powerful tool to make requests and store this huge mass of data (data lake). The most common solution is a Hadoop cluster, so you have to make sure that the flight data decoding solution you chose interfaces easily with Hadoop. In practical terms, you have to be able to configure the location of files once they’re decoded. The best tools even offer a PARQUET file format for your decoded data, a native format of a Hadoop cluster. If you’re not yet familiar with the Hadoop framework, it’s worth noting that one of its major advantages is its open source and open access configuration.

05 The next step

Whether you’re mainly concerned about flight safety, fuel costs or maintenance costs, if you meet these four criteria, you’ve already fulfilled all the conditions needed to optimize the analysis of your flight data. And the next step is up to you!

Reflecting our user-centric approach, every two years we organize a conference for all Cassiopée customers. This four-day event allows us to talk and work together, and address ways of making our products and services even more effective. Thanks to this road map we develop together, we can now offer you Cassiopée NODE. Not only does Cassiopée NODE address the four aforementioned points, but our solution also offers an extra advantage, namely a monitoring system that allows you to display progress on all decoding. But that’s not all! Since Cassiopée NODE is also designed to be integrated into a broader digital environment, you also have access to its API.

Together, let’s put your flight data to work!

Scroll to Top