Skip to main content
European Commission logo
español español
CORDIS - Resultados de investigaciones de la UE
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary

A Novel and Affordable Multi-Fidelity Deep Neural Network Uncertainty Quantification/Robust Optimization Design Framework for Industrial Turbomachinery

Descripción del proyecto

Un nuevo marco respalda el diseño de la turbomaquinaria multifásica más ecológica del mañana

La turbomaquinaria son equipos que transfieren energía a través de la expansión o compresión de un fluido en continuo movimiento a través de álabes rotadores, e incluye turbinas y compresores. La turbomaquinaria multifásica aprovecha varios ciclos de esta expansión o compresión de una serie para alcanzar un diferencial de presión muy elevado desde la entrada hasta la salida. El equipo del proyecto MENTOR, que cuenta con el apoyo de las Acciones Marie Skłodowska-Curie, está desarrollando un marco de red neuronal profunda que modelizará los efectos de incertidumbre sobre el rendimiento de la turbomaquinaria multifásica en las fases avanzadas del diseño de la turbomaquinaria multifásica. Las turbinas y compresores optimizados con mayores eficiencias contribuirán a los objetivos de neutralidad climática y de sostenibilidad de Europa.

Objetivo

"This fellowship aims to train a talented early career researcher and to contribute to the EU scientific excellence by developing an innovative Multi-fidelity dEep neural Network uncerTainty quantificatiOn and Robust optimization design (MENTOR) framework in order to handle the high dimensionality (HD) uncertain problems in the advanced multistage turbomachinery design process. The applicant is a highly dedicated and motivated young researcher and has been stimulated to propose this novel idea. He has been successively honoured with several prestigious awards including the National Scholarship for PhD Candidate, Excellent PhD Graduates of Beijing, Excellent Doctoral Dissertations Award of Beijing and ASME Young Engineer Turbo Expo Travel Award attributed to his excellent research achievement in cost-efficient uncertainty quantification (UQ) studies.
The traditional UQ methods can hardly control the computation cost for predicting the higher-order moments of multistage turbomachinery performance considering HD input uncertainties. The deep learning technology is a promising approximator in predicting the HD function. Integration of multi-fidelity (MF) methodology with deep neural network (DNN) can further combine their complementary merits. Thus, the novel MF-DNN method is proposed here and its effectiveness in handling a 60-dimensional test function has been preliminarily validated in the Incoming Researcher's recent work. Through this research fellowship, an affordable MENTOR framework will be finally established to investigate the multi-source uncertainty effects on multistage turbomachinery performance. This project has been carefully designed to match the applicant's profile with the strength of Imperial's UQ Lab, and thus will facilitate excellent two-way knowledge transfer and training activities. Successful completion of this fellowship will contribute to achieving the goal of the EU's ""Green Deal"" and will benefit the applicant's academic career prospect."

Coordinador

IMPERIAL COLLEGE OF SCIENCE TECHNOLOGY AND MEDICINE
Aportación neta de la UEn
€ 212 933,76
Dirección
SOUTH KENSINGTON CAMPUS EXHIBITION ROAD
SW7 2AZ LONDON
Reino Unido

Ver en el mapa

Región
London Inner London — West Westminster
Tipo de actividad
Higher or Secondary Education Establishments
Enlaces
Coste total
€ 212 933,76