Thiago Ritto, Associate Professor in the Dept. of Mechanical Engineering at the Federal University of Rio de Janeiro
Title: Uncertainty propagation through matrix perturbation and parameter identification using reinforcement learning applied to structural and rotor dynamics
Abstract:
In this talk, I will present some of the research I’m developing at the Federal University of Rio de Janeiro. I’m interested in dynamical systems (drill-strings, wind turbines, rotordynamics, fluid-structure interaction,…), predictive models (physics-based, machine learning, neural networks,…), uncertainties (stochastic analysis, robust design), and probabilistic learning (Bayesian framework).Then, I will expose two recent developments. The first one concerns a rotordynamic finite element model (ROSS). We evaluate how matrix perturbations propagate throughout different reduced-order models (modal reduction, Krylov subspace, Guyan reduction, and SEREP). The second one concerns a simple nonlinear oscillator. Experiments and simulations are used to propose a new strategy for model selection and parameter identification called RL-ABC (reinforced learning and approximate Bayesian computation). ABC is used to update the prior distribution and RL is used to speed up the model selection; models that perform better are reinforced having a higher probability of being selected again.
Bio
Thiago Ritto is an Associate Professor in the Dept. of Mechanical Engineering at the Federal University of Rio de Janeiro (UFRJ) since 2011 and a Research Fellow at the University of Bristol since 2021. He obtained his D.Sc. in Mechanical Engineering from Université Paris-Est (Lab MSME) and PUC-Rio in 2010. His teaching and research interests are related to dynamical systems (drill strings, wind turbines, rotordynamics, fluid-structure interaction), predictive models (physics-based, machine learning, neural networks), uncertainties (stochastic analysis, robust design), and probabilistic learning (Bayesian framework).https://thiagoritto.blogspot.com/
https://scholar.google.pt/citations?hl=fr&user=-qSNuJkAAAAJ
