PhD student - algorithms for detailed characterization and feature extraction (R1) per al Barcelona Supercomputing Center

The quest for carbon-neutral aviation propels us toward uncharted technological frontiers. In ERC Synergy TRANSDIFFUSE, we present an ambitious program to develop an AI-driven model that may revolutionize propulsion technologies. As a synergistic consortium, we unite the numerical modelling prowess of UPM team, the high-fidelity computational capabilities of LS/CFD group at BSC, and Purdue University experimental ingenuity.

TRANSDIFFUSE aims to develop new tools related to clean propulsion, specifically an AI-driven model called FluidGPT. In doing so, we expect to enable major advancements to the aeronautical and power generation sectors, including, for example, hydrogen pressure gain combustion (PGC) engines, which are compact, lightweight, high-efficiency turbines. In initiating FluidGPT, we aim to overcome challenges like those facing PGC, by exploiting the thing that has made compact turbomachinery so difficult to design: the troublesome transonic flows propelled from the combustor. With our model we aim to control and manipulate those transonic flows.

TRANSDIFFUSE will unfold in structured phases, beginning with extensive experimental and computational exploration of unsteady transonic turbine phenomena; followed by creation of FluidGPT, informed by the physics captured in Phase I; and last, applying FluidGPT to discover novel flow control strategies and exploit flow instabilities. Most of the tools required to achieve TRANSDIFFUSE are in the early stages of development. We intend to advance or develop theoretical modelling and computational and experimental tools as we progress through the project. Our collaborative endeavour thrives on the dynamic interplay between

computation, modelling, and experimental disciplines, ensuring that insights gleaned from each area amplify the impact and accuracy of the others.

The present PhD proposal is contained in the Phase I researching in algorithms for detailed characterization and feature extraction including development of methods to carry out eigenmode analysis taking advantage of novel GPU platforms.

Key Duties

• To develop feature extraction for massive data sets accelerated by means of GPU
• To assess the developed algorithms in TRANSDIFFUSE basic test cases
• To participate in the rest of the CFD activities of the LS/CFD team in CASE.
• To develop highly scalable eigenmode analysis algorithms

Data de tancament: Divendres, 31 Gener, 2025

Més informació

Més posts de Recerca