Reduced order modeling and machine-learning techniques for environmental flows problem

The project involves enhancing and extending the in-house fluid flow solver using advanced mathematical and computing framework to be undertaken at the host university. The current in-house solver code can be efficiently applied to aerospace, marine and environmental flow problems. The collaboration with host university will result in a more advanced version of the code due to the implementation of mathematics based reduced order models which will speed-up the computations without any loss of physics. The reduced order modeling method has proven to reduce the complexity of the problem from millions of elements to just tens or hundreds of elements with almost similar accuracy. Another part of the research collaboration consists of application of machine learning based models to the in-house flow solver. An artificial neural network will be fed in physics of the fluid flow problem and using training data it will be trained to “learn” how to solve the physics for the particular problem. These are the main objectives of the proposed research project which will be completed in the time frame of 24 weeks under the host supervisor.

Faculty Supervisor:

Artem Korobenko

Student:

Partner:

Scuola Internazionale Superiore di Studi Avanzati

Discipline:

Computer science

Sector:

Environmental Science and Technology; Artificial Intelligence

University:

University of Calgary

Program: