The Challenge
As with many embedded instruments, the speed calculation system in an aircraft is a critical element that cannot be allowed to fail for long. Pitot tubes are the heart of this system. In connection with an anemometer, their mission transcribes an aircraft’s relative speed by measuring the total pressure exerted in the sensor environment.
Even if Pitot tubes are positioned at points in the aircraft fuselage where airflow is disturbed as little as possible, temporary inconsistencies can sometimes occur in speed measurements taken due to partial or complete obstruction of the probe. At high altitudes or in areas of turbulence, the appearance of ice crystals is often the cause, whereas, on the ground, the fault may lie with simple insects.
To limit this type of risk as much as possible and further increase safety in its aircraft, a major European aircraft manufacturer has called on Expleo’s know-how to develop a backup speed calculation system capable of supplementing Pitot tube readings through a refined analysis of data sets.
Solutions
For six months and in two distinct phases depending on the aircraft engine type studied, Expleo set the task using the Python programming language or the artificial intelligence tools TensorFlow and Keras. Faced with many parameters to be considered, Expleo decided to work on a deep learning neural network model to better meet the customer’s integration requirements on embedded computers.
After first targeting some fifteen critical parameters in calculating an aircraft’s speed, Expleo selected the three most relevant, namely pressure sensors (total and static) from engines and the rotation speed of the blades.
By screening them with the neural network model developed for the occasion, experts in charge of this project established correct speed measurements in case of various disturbances with a margin of error of only 10 knots (i.e. about 11 mph) on Pitot tubes.
Outcome
Of the two engine types studied, the models developed have been validated by our customer’s design office, and the third project on a new engine type is already planned. Even if these different speed calculation backup systems have not yet been installed in the aircraft concerned, promises seen in the accuracy of our models are already delighting the European aircraft manufacturer, who is the master of this project.
With such systems, our customer’s aircraft will benefit from better robustness and accuracy in their flight plans by avoiding the consequences of all those temporary shifts that can sometimes occur in the readings taken at high altitudes.
With this project’s success, our customer again recognises Expleo's strong aeronautical engineering and data science expertise. If a delta is recorded between the speed taken in their aircraft and the mathematical model we created, sensor anomalies are automatically detected and reported to pilots. This is a real bonus in case of possible Pitot tube obstructions.
DIGITAL TRANSFORMATION
Big data, analytics, AI & advanced algorithms
Expleo builds modern data analytics strategies by deploying machine learning and advanced algorithms. This empowers us to create complex predictive and governing models, both in the industry and services environments.
