We look at the evolution of flight data monitoring and whats just around the corner By First Officer Adam CrehanClark, BALPA member s it a big brother looking out for you in the playground? Or Orwells dystopian Big Brother watching your every move? Hopefully, by now, most pilots will have an opinion on flight data monitoring (FDM), leaning more towards the former. FDM has its roots in flight data recording for crash purposes, dating back almost as far as commercial flight itself. The earliest flight data recorder (FDR), in the 1930s, was a French design by Franois Hussenot and Paul Beaudouin. This used scrolling photographic film to record deviations of light reflected from a mirror, from which very basic data such as speed or altitude could be inferred. This early FDR was used in early flight testing and in data recording. The more familiar type of variable recording began a few years later, during World War II. Len Harrison and Vic Husband created a crash-proof FDR that involved a stylus physically connected to a sensor (altimeter or airspeed) that would move with the sensor and draw lines on revolving copper foil. The whole unit was housed in a crash-proof floating buoy and formed the early flight data recorder colloquially known as a black box. This name derives from the fact that FDRs, plus other secret RAF electronic units including radar and weapons were housed in secretive black boxes to be placed on aircraft during the war. Now the boxes are coated with heatproof, flame-orange paint so they can be seen among wreckage. The step from recording flight data in case of an incident to recording flight data for practical and monitoring purposes came in the 1960s, with the advent of the Quick Access Recorder (QAR). The QAR is a smaller unit, not designed to withstand a crash but, as a trade-off, it can record more data, at a much higher rate than an FDR. British European Airways (BEA) installed these units onto their Hawker Siddeley Trident to prove to the CAA that the aircrafts autoland system was safe enough for operational use. From here, manufacturers, airlines and regulators have developed the data-recording and playback systems into what we now see and recognise as FDM, also known as flight operations quality assurance (FOQA) in the US. Over the past five decades since BEAs autoland testing, the technical, regulatory and practical areas of FDM have come on in leaps and bounds. Working it out How does modern FDM work? First, we need to understand some basic machine architecture. Todays aircraft are largely digital machines and, like most modern transport machines, they use a databus to relay information around the vehicle. In much the same way as the human bodys central nervous system or spinal cord connects the brain to its senses (input) and muscles (output), the databus links a machines sensors (inputs) to its centralised computer (brain) and back around to its servos, motors or lights (outputs). Every few milliseconds, the databus sends sensor positions, in binary, to the centralised computer, which compares the previous position of the sensors and decides what the output positions should be. The computer replies by sending these new positions back out on the databus in one long binary stream, for the individual components to pick up. For example, if the flap switch position (input) has moved from the previous cycle to this cycle, the flap motor (output) needs to be activated. The centralised computer will send a message on the databus and, when the unit that controls the flaps hears the command, it will begin the movement process. This means that, essentially, there are only two binary messages being sent along the databus at any time the sensor input message to the centralised computer and the output message from the centralised computer to the output controllers. A flight data acquisition unit (FDAU) listens in to the databus and translates the code from binary to something that the FDR and the QAR can understand and record. This is often a translation to hex code, which is easier to store. Because of memory limitations, both units pick and choose which information they want to record off the databus. Older models had a handful of parameters they could record to memory cards, which had to be physically downloaded after each flight, or sequence of flights. Newer models can record virtually every input and output data point, and then automatically send the data via the mobile network or wireless internet to the main office. Boeing has trialled a new system on the 777F that relays back FDR flight data in real time via satellites to a main office. There has been a particular push for this kind of real-time awareness in the wake of the MH370 disappearance. Once the FDM is downloaded from the aircraft, it is reconciled through a software platform that allows the user to not only read the data, but also recreate the flight. To do this, the software must translate the hex code into something usable by a human. This process is often time-consuming, because the computer needs to understand which bit of data relates to which sensor or output. So huge reference tables need to be built and maintained, especially for the newer, all-encompassing systems. Considering that some operations do many hundreds of sectors per day, it makes for a lot of number crunching! Once this is complete, the hard part is turning the data into something meaningful that can lead to a safer operation. TODAYS AIRCRAFT ARE LARGELY DIGITAL MACHINES AND USE A DATABUS TO RELAY INFORMATION AROUND THE VEHICLE Looking at trends We live in a largely reactive FDM world, where sensor-value exceedances trigger events of varying severity. For example, the nearly inevitable and pride-damaging hard landing will be listed as a low, medium or high event based upon the G value sensed at touchdown, often set well below the manufacturers limits. The airlines safety management system (SMS) can use these events to work out if they are trending up or down over time, and if there is any operational change that could explain the trend. As Im sure many of us have seen, the SMS can replay a 3D version of the flight to crews in slow motion, to aid memory or to highlight some situational awareness that may have been missed. Very often, the identities of individual pilots are encrypted in the FDM, and BALPA plays a crucial role in being the guardians of that encryption, allowing a double layer of data protection for crews. Alongside this, some SMS allow pilots to review their own flights in their own time, to help with debriefing or learning points. After all, it is your data. But whats next? On the horizon is a system that can use big data to predict or highlight when the next event is potentially going to happen. This crystal ballstyle future is a tall order, however the technological hurdles are huge, not just in manipulating the data, but also acquiring and storing it. The two main projects, either side of the Atlantic, tackling this problem are Eurocontrols Data4Safety (D4S) and the US governments Aviation Safety Information Analysis and Sharing (ASIAS). Both intend to marry data from FDM, air safety reports, weather, ATC reports, radar, flight planning and airport/ airspace infrastructure into huge, multifaceted datasets that can be mined in more advanced ways to enhance safety across the industry. Currently, both are in proofof-concept phases. Their key objective is to prove themselves trusted guardians of the data, ensuring it is protected and used solely for safety. Both parties have buy-in from airlines, air navigation service providers, regulatory authorities, weather providers, charting and geographic companies, governments, and airport authorities across their relevant regions. In a few years, we should have whats known as a data observatory, where we can tell what a normal flight is for almost all situations, based upon the data. We can then teach a computer algorithm what a normal flight is for a particular set of conditions. This machine-learning algorithm can be used to find the non-normal flights, and tie up all the external data that contributed to the event such as specific weather, airspace traffic, aircraft type, time, location and delays. Then, by working backwards when those specific conditions are predicted to occur again, the computer can warn us and, potentially, turn a non-normal flight into a normal one. The results remain to be seen, but hopefully, sometime soon, this work will come to fruition and make a tricky day out just a bit easier. TECH LOG THE PAST, PRESENT AND FUTURE OF FLIGHT DATA MONITORING