NASA Meets With Predii to Discuss Applied Data Science
We met with NASA's Dr. Nikunj Oza to discuss data science projects and collaboration.
The Predii team met last week with Dr. Nikunj Oza, NASA’s Data Science Group Lead, to discuss ongoing projects his team is working on. Predii has a history with NASA – it has been collaborating with the Unmanned Aerial Vehicles Collaborative, exploring applications in safety and operations.
Dr. Oza’s team started as a Machine Learning R&D group back in 2002, and it tackles Data Science problems across the general NASA organization. They have been applying machine learning to NASA fields such as Aeronautics, Earth Science, Space Exploration and Space Science, regularly publishing contributions to the research literature. During this meeting with Predii, significant key talking points were projects related to commercial aviation.
One of the key discovery points for the NASA Data Science Group centers around statistically anomalous flights, with a focus on assessing and diagnosing precursor events to anomalous flight states. Their goal is to assess flight safety, and power improvements to safety operations and procedures with insights gained from analyzing various flight data.
The Safety Life Cycle for aviation follows a four step, recursive process: Identify, Evaluate, Formulate, and Implement.
Dr. Oza’s team primarily focuses on two steps: Identify and Evaluate. Aviation already has an extremely extensive depth of understanding for flight safety causes and corrections, and has a similarly extensive set of procedures for implementing any changes to flight operations – for good reason. Dr. Oza’s team builds its most significant value in detecting and evaluating events that lie outside of the norm. Domain experts then take his team’s findings and formulate their implementation.
This value primarily derives from the current method of creating aviation safety procedures, which are largely Exceedance-Based: once you exceed the parameters set up in the rules, you get flagged. This rules-based methodology has created a fantastically safe flight safety system – by far the safest method of traveling – but it is not as agile at anomaly detection. It has a “low false positive rate, high false negative rate (missed detection) rate.”
Integration NASA’s data driven methods into current Exceedance-Based methodology is the best of both worlds.
(Please note that the visualization above is not to scale. The Anomalous portions would not be visible if they were to scale.)
Dr. Oza’s team finds the Unknown Problems, evaluates the sequence of events that cause them, and relays their findings to flight safety professionals, contributing to the flight safety improvement feedback loop.
NASA's UAV & Predii have engaged in mutual collaboration with Dr. Nikunj Oza (Leader, Data Sciences Group) in sharing our findings from industry via Seminar talks. We plan to multiple seminar talks in future and continue collaboration.
For more information on Dr. Nikunj Oza, view his profile on the NASA website here.