Banca de DEFESA: João Paulo Silva Lima

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : João Paulo Silva Lima
DATE: 07/07/2023
TIME: 09:00
LOCAL: Plataforma Teams
TITLE:

BI-FIDELITY SURROGATE MODELING WITH SUBSET SIMULATION FOR STRUCTURAL RELIABILITY ASSESSMENT

KEY WORDS:

Reliability; Artificial Neural Networks; Multi-task; Multi-fidelity; Non-linear Finite Element Analysis

PAGES: 107
BIG AREA: Engenharias
AREA: Engenharia Civil
SUBÁREA: Estruturas
SPECIALTY: Mecânica das Estruturas
SUMMARY:

This thesis presents a novel Bi-Fidelity Multi-task Learning Model based on a Deep Neural Network (BFMT-DNN) to address the computational challenge of structural reliability analysis applied to complex structures. The main contribution is the development of a novel hyperparameter-optimized BFMT-DNN, that considers the advantages of Bayesian Optimization, focusing on prediction accuracy, stability, and computational efficiency to accesses the reliability of high nonlinear problems. For constructing this surrogate model, the study proceeded throughout two stages before. Firstly, is presented a method based on a Bi-fidelity Kriging surrogate model associated with Subset Simulation for structural reliability analysis. The efficiency of the bi-fidelity Kriging model is evaluated using a stiffened panel reliability problem that demands high computational costs, such as non-linear finite element analysis structural models. The next step proposed a two-stage Bi-Fidelity Deep Neural Network surrogate model in association with Subset Simulation to quantify the uncertainty of structural analysis and assess the probability of failure of high dimensional rare events. In the two steps, the surrogate models can reproduce the non-linear behaviour in the variable's uncertainty analysis, reducing the high computational demand of these problems. Furthermore, the BF-DNN surrogate model used Bayesian optimization to fine-tunning the hyperparameters. The multi-fidelity models used low-fidelity data samples added to the model to predict high-fidelity responses, and, when presenting a good correlation between the fidelities, the assessment of the proposed method showed that the proposed Multi-fidelity Method is a good strategy because it can provide an accurate probability of failure estimation with a lower computational cost. A hyperparameter-optimized BFMT-DNN using low-fidelity data samples added to the model to predict high-fidelity responses for structural collapse behaviour framework is presented in the final analysis. The assessment is realized in an offshore wind turbine in extreme conditions and described using non-linear Finite Element analysis to obtain multiple outputs. The results show that the proposed multi-fidelity methods can give a precise failure probability estimation with less computational cost.

COMMITTEE MEMBERS:
Presidente - 2047809 - FRANCISCO EVANGELISTA JUNIOR
Interno - 404840 - WILLIAM TAYLOR MATIAS SILVA
Externo ao Programa - 404227 - MARCIO MUNIZ DE FARIAS - UnBExterno à Instituição - CARLOS ANTÓNIO PANCADA GUEDES SOARES
Externo à Instituição - MAURO DE VASCONCELLOS REAL - FURG