Flight and Passenger Safety (aviation safety) are crucial elements in the aviation industry. The Safety Management System is a specific department that looks after overall issues and anomalies that can be risky for aviation operations.
Data Analysis and aggregation can be used to mitigate the risk involved in Flight Operations, Navigation, Alert, and Maintenance.
Nowadays airplanes generated a lot of data about the engine systems, fuel usage, crew activity, and even weather systems that the crews encountered.
Today’s advanced jets
Today, through thousands of sensors and sophisticated digitized systems, the newest generation of jets collect exponentially more information. Each flight generates more than 30 times the amount of data that the previous generation of wide-bodied jets produced.
Only about one‑tenth of the current global aircraft fleet is made up of these technologically advanced aircraft, In a decade more than half of the fleet will be technologically advanced.
By 2026, annual data generation should reach 98 billion gigabytes, or 98 million terabytes, according to a 2016 estimate by Oliver Wyman.
The newest generation aircraft by then will be spewing out between five and eight terabytes per flight, up to 80 times what older planes generate today!
Airlines and airports still have only limited capacity to process this drove of information and use advanced analytics and artificial intelligence to help inform operations and maintenance. This data, in turn, is almost never transmitted in real time.
Discussing the connectivity of a plane rather revolves around whether passengers can get WiFi signals that let them do their work or stream their favorite entertainment.
How does this new data analytics technology work?
The prediction of anomalies or adverse events for aircraft security is a challenging task. There are a variety of methods which can be used to address that.
For that, we develop an algorithm which helps to find/detect anomalies during flight. The solution is based on the previous dataset.
After detection, it will generate an alarm to assist Air Traffic Control (ATC), Airline Safety Departments, Safety Authorities, Safety Associations, Air Insurance Companies, Aircraft Manufacturers, and Aviation Allied Industries across the globe.
This approach and method are currently in the Research phase of NASA. Also, our system can generate reports on specific parameters such as destination, source, altitude, speed, and location.
Operational cost savings for the airlines
The new connectivity and advanced analytics also mean savings for airlines. Oliver Wyman’s savings estimate is between 2 percent and 2.5 percent of total global operating costs. That translates to something between $5 billion and $6 billion of savings annually.
While it may take several years, even a decade, to realize all these possibilities, aviation is on the cusp of a data science revolution. That revolution will transform almost every aspect of the industry and provide its managers with better control.
The question industry players must ask themselves is whether they will be leaders or followers.
Airlines face daily unpredictability challenges
Artificial intelligence and advanced analytics can play the biggest role are dealing with the unpredictability the industry faces daily. With hundreds of planes, thousands of flights, and millions of employees and passengers, there is now too much data coming in.
There are too many variables for humans to sort through fast enough to fix problems or even prioritize potential threats.
Computers and analytics are necessary for big events such as hurricanes, snowstorms, air traffic control delays, mechanical failures, or even lines of thunderstorms. While much of this activity today is mostly reactive, the next step will be for aviation to proactively avoid some of the problems.
Such problems include delays, congestion, and inefficiencies that annoy passengers and keep the global industry at single-digit profit margins.
NASA’s role in the project
Currently only NASA is working on this project (to detect anomalies in spacecraft for its security). If there is any private in-house research going on it is not on the surface as yet.
No work has been done in this area for airplanes. NASA has worked on anomaly detection and prediction for space crafts but this is still unknown for aircraft. By doing research and studying NASA’s papers we have come to the point that NASA’s project on anomaly detection named ACCEPT’s objectives is to develop a framework to test adverse event prediction algorithms.
This is explicitly geared to providing a comparative performance assessment of results from the application of a variety of algorithmic methods. This type of toolkit is also needed for aircraft to ensure their safety.
Aviation safety means the state of an aviation system or organization in which risks associated with aviation activities, are reduced and controlled to an acceptable level. It encompasses the theory, practice, investigation, and categorization of flight failures. It also encompasses the prevention of such failures through regulation, education, and training.
Our system will reduce the risks related to aircraft by performing data analysis on large datasets made available via ADSB feeder. It will recognize patterns that when anomalies do occur.
- What the normal behaviors are, exhibited by the aircraft in terms of its g-force while turning.
- Altitude, when it encounters any turbulence.
- Go-arounds, and other dynamic occurrences. etc.
- Whenever the aircraft detects an anomaly or unusual behavior it will communicate such behavior to Air Traffic Control (ATC).
- It will also keep check of the proximities of aircraft and will make ATC aware if the current positions of aircraft can lead to a near future collision so ATC can make pilots take preventive measures such as a change in altitude or direction to avoid a possible collision.
Work to be done
We have to design a toolbox to detect anomalies And we have to work on several algorithms to develop this very toolbox. We will mainly be working on two main algorithms of
1) ML, e.g. regression algorithms for prediction/detection of anomalies, and
2) classification algorithms for activating alerts.
The major task of the toolbox is to detect the anomaly and after detection, it will generate an alert which assists ATC.
The alerts that the system generates might be false, depending on accuracy (false alert), and it might miss some anomalies too (missed alert). We will work to make this system as accurate as possible.
Our Toolbox detects anomaly in a way that it will check current data like latitude, longitude, and speed, etc and matches it with a stored dataset.
If there is any disturbance found in the data, or if the values are not matched, then our toolbox detects the current situation of aircraft as an anomaly and generates an alert.
Algorithms will be needed for working on the filtering of datasets after filtration, comparison of predicted value by toolbox and the actual value of the aircraft taking place.
The results of the comparison algorithm will be used as inputs for the classification algorithm. We will finalize the algorithms to be deployed after doing further research work.
Motivation and needs
There are various areas in which data mining can be used in the aviation sector to improve flight safety and to reduce disasters. Nowadays millions of lives are at potential risk, due to unpredictable accidents that occur during an aircraft’s flight operation.
Flight safety is currently being done by ATC to some extent, but it is really tough for ATC to keep an eye on everything at any point in time.
Our system will solve this problem-challenge of air traffic controllers and assist them in recognizing when anomalies occur.
The tasks of ATC are tough as they have to keep an eye on several aircraft at the same time. They’re responsible for the aircraft as it travels, using radar to track its exact position, keeping the airspace safe.
The job of ATC can be made easy if a toolkit was provided to them. The toolkit would analyze the real-time data and detect the anomalies themselves. Then the system would alert the air traffic controller of anomalies during dynamic flight operations.
Federal Aviation Administration (FAA)
Currently, the FAA is working on its NEXTGEN program. NEXTGEN is the modernization of the airspace for radar-based surveillance to precision satellite monitoring of all aircraft in our skies.
Its goal is to increase the safety, efficiency, capacity, predictability, and resiliency of American aviation.
This needed overhaul brings together innovative technologies, capabilities, and procedures that improve how we fly from departure to arrival.
The FAA has not yet worked on this anomaly detection and prediction technology. This made us motivated to work on this project as it will be a great project to attract the FAA. This, in turn, can be a huge contribution to their NEXTGEN program.
To be accomplished
Alpha AI Inc is working on the datasets of aircraft, provided by an External Advisor via the ADS-B feeder. The data itself, however, does not contain any clear indication of the anomalies that occur during flights.
Our team intends to:
- Develop algorithms to detect anomalies that flights may involve, including aircraft collision alerts, go-around situation, altitude dropping due to turbulence, anomalous G-Forces exhibited by the aircraft, proximity and speed anomalies.
- Implement a system that is capable of predicting any such anomalous behavior exhibited by aircraft, by analyzing the data that can lead to an incident.
- Determine the threshold values for altitude, g-force, speed, proximity, etc, for turbulence, turns, collisions, and turbulence.
- Design a system that will be able to assist Air Traffic Controllers. Controllers can be notified through a buzzer, by taking preventive measures against the anomalous behavior of the aircraft. This, in turn, can save lives.
- Develop a system that will be able to generate reports indicating anomalies.