Big Data in Aerospace

“Big data” is all abuzz in the media these days. As more and more people are connected to the internet and sensors become ubiquitous parts of daily hardware an unprecedented amount of information is being produced. Some analysts project 40% growth in data over the next decade, which means that in a decade 30 times the amount of data will be produced than today. Given this this trend, what are the implications for the aerospace industry?

Fundamentally, big data is nothing new for the aerospace industry. Sensors have been collecting data on aircraft for years ranging from binary data such as speed, altitude and stability of the aircraft during flight, to damage and crack growth progression at service intervals. The authorities and parties involved have done an incredible job at using routine data and data gathered from failures to raise safety standards.

What exactly does “big data” mean? Big data is characterised by a data stream that is high in volume, high velocity and coming from multiple sources and in a variety of forms. This combination of factors makes analysing and interpreting data via a live stream incredibly difficult, but such a capability is exactly what is needed in the aerospace environment. For example, structural health monitoring has received a lot of attention within research institutes because an internal sensory system that provides information about the real stresses and strains within a structure could improve prognostics about the “health” of a part and indicate when service intervals and replacements are needed. Such a system could look at the usage data of an aircraft and predict when a component needs replacing. For example, the likelihood that a part will fail could be translating into an associated repair that is the best compromise in terms of safety and cost. Furthermore, the information can be fed back to the structural engineers to improve the design for future aircraft. Ideally you want to replicate the way the nervous system uses pain to signal damage within the body and then trigger a remedy. Even though structural health monitoring systems are feasible today, analysing the data stream in real time and providing diagnostics and prognostics remains a challenge.

Other areas within aerospace that will greatly benefit from insights gleaned from data streams are cyber security, understanding automation and the human-machine interaction, aircraft under different weather and traffic situations and supply chain management. Big data could also serve as the underlying structure that establishies autonomous aircraft on a wide scale. Finally, big data opens the door for a new type of adaptive design in which data from sensors are used to describe the characteristics of a specific outcome, and a design is then iterated until the desired and actual data match. This is very much an evolutionary, trail-and-error approach that will be invaluable for highly complex systems where cause and effect are not easily correlated and deterministic approaches are not possible. For example, a research team may define some general, not well defined hypothesis about a future design or system they are trying to understand, and then use data analytics to explore the available solutions and come up with initial insights into the governing factors of a system. In this case it is imperative to fail quickly and find out out what works and what does not. The algorithm can then be refined iteratively by using the expertise of an engineer to point the computer in the right direction.

Thus, the main goal is to turn data into useful, actionable knowledge. For example in the 1990’s very limited data existed in terms of understanding the airport taxi-way structure. Today we have the opposite situation in that we have more data than we can actually use. Furthermore, not only the quantity but also quality of data is increasing rapidly such that computer scientists are able to design more detailed models to describe the underlying physics of complex systems. When converting data to actionable information one challenge is how to account for as much of the data as possible before reaching a conclusion. Thus, a high velocity, high volume and diverse data stream may not be the most important characteristic for data analytics. Rather it is more important that the data be relevant, complete and measurable. Therefore good insights can also be gleaned from smaller data if the data analytics is powerful.

While aerospace is neither search nor social media, big data is incredibly important because the underlying stream from distributed data systems on aircraft or weather data systems can be aggregated and analysed in consonance to create new insights for safety. Thus, in the aerospace industry the major value drivers will be data analytics and data science, which will allow engineers and scientists to combine datasets in new ways and gain insights from complex systems that are hard to analyse deterministically. The major challenge is how to upscale the current systems into a new era where the information system is the foundation of the entire aerospace environment. In this manner data science will transform into a fundamental pillar of aerospace engineering, alongside the classical foundations such as propulsion, structures, control and aerodynamics.