Airbus
Smart Surrogate Modelling for efficient, correlated Time History prediction based upon a very reduced set of aircraft parameters (SSMTH)
Description of the company
Airbus is a global pioneer in the aerospace industry, operating in the commercial aircraft, helicopters, defense and space sectors. The Company is a leader in designing, manufacturing and delivering aerospace products, services and solutions to customers on a worldwide scale. Hamburg, the headquarters of Airbus Commercial Aircraft in Germany and the largest Airbus site in Germany, employs over 15.000 people and plays a key role in the development and manufacturing of all Airbus aircraft.
In the engineering domain, key competences for overall aircraft design in Hamburg involve “Mass Properties” and “Loads and Aeroelastics” for all Airbus aircraft in-service and under development.
Situation
Surrogate modelling has become state-of-the-art in many areas of daily life today.
Problem
However, technical applications of surrogate modelling for which safety represents the essential criterion are still rarely deployed, even though many use cases exist. An additional challenge represents the requirement to output correlated time histories i.e. , represent the interdependencies of system outputs with an efficient architecture of a surrogate and related machine learning.
Aims of the project
In-service data is valuable to understand the usage of the product in operation. Whereas data from e.g. FlightRadar24 cover information being recorded from “a ground station” e.g. rudimentary trajectory related information, onboard recordings deliver a more comprehensive view on aircraft parameters on a flight-by-flight basis. Still some information needs to be recovered based upon correlations that are given by available parameters.
Here, complementary time history information shall be generated via a machine learning model. Based upon public in-service recordings containing e.g., the angle of attack, altitude, mach number, control surface deflections vs time, the surrogate shall deliver e.g., a “correlated” vertical load factor as time history.
Scopes
The following core items of interest need to be addressed:
1. Explore solutions from industries with related requirements / application context
2. What is recent (scientific) state-of-the-art, here?
3. What is an adequate, necessary sufficient level complexity of the surrogate for capturing /delivering correlated time histories adequately?
4. Extract and define necessary and sufficient design criteria
a. Technical performance criteria
b. Consider particularities related to safety relevant applications (EASA Artificial Intelligence concept paper - proposed Issue 2)
c. Quantify the resulting uncertainty of the surrogate output
d. Quantify the explainability of the surrogate output
5. Discuss options to fulfil the design criteria and related methods as e.g. Fast Fourier Transform (FFT), Principal Component Analysis (PCA), Autoencoder, neural network with Long Short-Term Memory cells (LSTM), Subspace Models
6. Develop and test alternative solutions
7. Perform validation and verification
8. Pitch your solution
Target group (students)
Master´s Programs (6 participants - preferably with at least 2 different programs) with related context
● Electrical & Mechanical Engineering, Aeronautics
● Informatics, Mechatronics, Data Analysics
● Microelectronics and Microsystems
● Product Development, Materials and Production
● International Management and Engineering
Dates
Please save these dates: Fishing for Experience Termine
Registration
You can apply for Fishing for Experience online.
