MACHINE LEARNING APPLIED TO FLOWS: EXPERIMENTAL AND NUMERICAL ANALYSIS

• Grant number: 24/08030-0
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: August 01, 2024
• End date: April 30, 2026
• Field of knowledge: Engineering - Mechanical Engineering - Transport Phenomena
• Agreement: BG E&P Brasil (Shell Group)
• Principal Investigator: Emílio Carlos Nelli Silva
• Grantee: Vitor Augusto Andreghetto Bortolin
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Company: Universidade de São Paulo (USP). Escola Politécnica (EP)
• Associated research grant: 20/15230-5 - Research Centre for Greenhouse Gas Innovation - RCG2I, AP.PCPE

Abstract

Recent advances in computational and measurement techniques have enabled an increasingly detailed study of fluids. However, this exponential growth in data volume presents major challenges. The first of these is the extraction of conclusions and relevant information from this vast set of data. Furthermore, experimental cases often suffer from strong noise when advanced techniques are employed in an attempt to obtain maximum accuracy, as is the case with particle image velocimetry (PIV). In this context, machine learning brings new tools and methodologies that allow analysing, synthesizing and filtering data efficiently, enabling significant advances based on the data obtained. This work proposes the development and application of machine learning methodologies to results obtained both experimentally and numerically in fluid mechanics. PIV measurements will be the source of experimental data, while computational fluid mechanics simulations (CFD) will be the source of numerical data. Initially, the project focuses on unsupervised machine learning methods, in particular the POD decomposition, which allows the identification of flow macrostructures, the RPOD, which allows the filtering of heavily corrupted experimental results, and the t-SNE, which is a powerful synthesis tool for large volumes of data. Therefore, this project also aims to significantly expand our understanding of the behaviour of these complex fluid systems.