Turning Raw Data into Models to Improve Aviation Fuel Efficiency

Published Apr-16-18

Breakthrough:
Modeling fuel consumption by airplanes to gain fresh insights into how to increase fuel efficiency, which would drive down costs and benefit the environment.

Company:
Honeywell International Inc., United States

The Story:

Turning Raw Data into Models to Improve Aviation Fuel Efficiency Aviation fuel is expensive, contributing to about 30% of the operating costs of airlines. It also has a considerable carbon footprint. According to the Center for Climate and Energy Solutions, aviation is responsible for 1.5% of the global anthropogenic greenhouse gas emissions. Consequently, many carriers across the world are engaged in efforts to go green which should have a significant positive impact on the environment. One way of achieving this is through reducing fuel consumption which would also save a lot of money.

To glean insights on how to make flights more fuel-efficient Honeywell International Inc., an American multinational conglomerate company held a data analytics contest on the CrowdAnalytix platform.

Predicting Fuel Flow Rates

Participants were tasked with predicting the fuel flow rate of airplanes during phases of a flight, including, taxiing, take off, climbing, cruising and approach. The aim was to investigate the data to understand drivers of fuel flow (consumption) and come up with best practices that make flights fuel efficient under different conditions.

To help data analysts in their pursuit Honeywell released several files of flight data that contained readings of numerous sensors on planes, covering aircraft dynamics and system performance among other parameters.

Participants looked at training and test datasets to develop their models which were not only to be used for making predictions. Honeywell wanted predictive models that could be utilized in innovative ways to identify key insights.

Investigating the Raw Data

Among the set of tasks solvers had to perform in the first phase of the contest were to build their model from the training data set and derive unique features based on the raw data (such as the number of times brakes were applied and the number of turns a plane made while taxying).

Eight conditional winners were then chosen from the private leaderboard and these went forward to the contest's next stage in which submissions were further evaluated based on additional criteria, such as fuel flow predictions during various stages of the flight.

Winners

The contest attracted 362 solvers and the eight conditional winners shared a prize pool of $7,500. In first place was Stijn Tilborghs from Cracow in Poland who picked up a check for $2,500. The award for second place and a check for $1,500 went to deorani from Singapore, while Pietro Marini from Paris, France came in third and was given a prize of $1,500. Five consolation prizes of $500 each were given to the other five winners.

Share on        
Next Story »

What Can we Solve for You?
Airplane_jet_engine_,_Водухопловен_млазен_мотор.JPG