Banca de DEFESA: DIEGO ZILLI LIMA
Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.DISCENTE : DIEGO ZILLI LIMA
Data: 17/06/2026
HORA: 14:00
LOCAL: https://meet.google.com/cmc-pnub-eqe
TÍTULO:
A Constrained Bayesian Optimization Framework for Resonance-Aware Design of Centrifugal Compressors Based on SAFE Diagrams
PÁGINAS: 165
RESUMO:
The structural design of centrifugal compressors operating under demanding conditions, such as those encountered in supercritical CO2 (sCO2) systems, requires the simultaneous consideration of mechanical integrity and dynamic performance. In particular, resonance phenomena play a critical role in the reliability of these machines, as they may lead to excessive vibration levels and high-cycle fatigue failures. Despite their importance, resonance conditions are typically treated as a post-design verification criterion and are not explicitly incorporated into the optimization process.
This work proposes a novel framework for resonance-aware structural optimization of centrifugal compressors based on Constrained Bayesian Optimization (CBO). The main contribution lies in the reformulation of resonance as a continuous and optimization-ready objective function. A scalar metric derived from the SAFE (Singh Advanced Frequency Evaluation) diagram is developed, enabling the quantification of resonance risk by simultaneously accounting for frequency coincidence and nodal diameter compatibility. This approach overcomes the limitations of traditional Campbell-based methods, which consider only frequency proximity and may identify non-physical resonance conditions.
The proposed framework integrates high-fidelity nonlinear finite element analyses, pre-stressed modal analysis, and aerodynamic loading through a one-way fluid-structure interaction approach within a probabilistic optimization environment. In addition, new constraint-handling strategies are introduced through the development of weighted acquisition functions, combining weighted Constrained Expected Improvement (wCEI) and weighted Probability of Feasibility (wPoF). These strategies enhance the balance between feasibility exploration and optimality in highly constrained design spaces.
A comprehensive validation study using analytical benchmark problems demonstrates the robustness and effectiveness of the proposed methods. The results show that probabilistic approaches outperform classical penalty-based formulations, with CEI providing improved convergence stability and wCEI enhancing early feasibility identification.
The methodology is applied to the structural optimization of a centrifugal compressor impeller operating with supercritical CO2. The optimized designs exhibit reduced resonance risk, lower stress levels, and full compliance with structural constraints. In particular, the SAFE-based objective function successfully eliminates critical resonance crossings identified in the baseline design.
Overall, this work advances the state of the art by bridging resonance physics and probabilistic optimization, providing a robust and computationally efficient framework for the design of turbomachinery components under realistic operating conditions.
MEMBROS DA BANCA:
Presidente - Interno ao Programa - 2312727 - DANIEL JONAS DEZAN
Membro Titular - Examinador(a) Externo ao Programa - 2278416 - MARCELO ARAUJO DA SILVA
Membro Titular - Examinador(a) Externo à Instituição - RODRIGO LIMA STOETERAU - USP
Membro Titular - Examinador(a) Externo à Instituição - RENATO PICELLI SANCHES - USP
Membro Titular - Examinador(a) Externo à Instituição - RODRIGO NICOLETTI - USP
Membro Suplente - Examinador(a) Interno ao Programa - 1985515 - ANTONIO GARRIDO GALLEGO
Membro Suplente - Examinador(a) Externo ao Programa - 3042266 - WALLACE GUSMAO FERREIRA
Membro Suplente - Examinador(a) Externo à Instituição - RODRIGO VIDONSCKY PINTO - IFSP