Ofertes treballs fi d'estudis

Títol: Machine-learning surrogate modeling of nonlinear lattice structures

Director/a: Yago Llamas, Daniel

Email del professor/a: daniel.yago@upc.edu

Departament del professor/a: 748-FIS

Codirector/a: Cante Teran, Juan carlos

Email del codirector/a: juan.cante@upc.edu

Departament del professor/a: 748-FIS

• 2010 - GRAU EN ENGINYERIA EN TECNOLOGIES AEROESPACIALS
• 2010 - GRAU EN ENGINYERIA EN VEHICLES AEROESPACIALS

Identificador de l'oferta: 205-06439

Modalitat: Universitat

Possibilitat de beca/finançament: No

Paraules clau:

Descripció:

The objective of this thesis is to analyze the nonlinear mechanical response of lattice-based structures using finite element simulations and to develop machine-learning surrogate models that accurately predict their behavior. The project focuses on accelerating nonlinear FEA by creating fast-inference models capable of reproducing the macroscopic response of the lattice, including energy dissipation mechanisms driven by micro-buckling. Additionally, the thesis aims to study how variations in lattice topology and beam-level properties influence global performance. The final outcome is a validated ML surrogate capable of assessing these effects efficiently.

Objectius:

The objectives of this thesis are to study the nonlinear mechanical behavior of lattice-based structures and to develop machine-learning surrogate models capable of predicting their response efficiently. The work begins with a literature review on lattice architectures, nonlinear deformation mechanisms—especially micro-buckling—and machine-learning methods suitable for regression of structural responses. Finite Element Analysis (FEA) will be used to simulate representative lattice cells and extract their macroscopic behavior. A surrogate neural network model will then be trained to reproduce the nonlinear response and energy characteristics based on topology parameters or beam properties. Finally, the thesis aims to evaluate how these changes affect global performance and to summarize all findings in a comprehensive final report.

Tasques a realitzar / Característiques:

-Literature Review: Perform a comprehensive review of lattice-based structures, nonlinear deformation mechanisms such as micro-buckling, and existing numerical approaches for modeling their response. Additionally, study relevant machine-learning methods—especially neural networks—used for surrogate modeling of complex mechanical systems. -Finite Element Model Development: Implement and validate nonlinear Finite Element Analysis (FEA) models of representative lattice (beam-based) unit cells. Simulate their mechanical response under different loading conditions to extract force–displacement behavior and energy dissipation at the macro-scale. -Dataset Generation: Systematically vary geometric and material parameters of the lattice microstructure, run FEA simulations for each configuration, and compile a clean and well-organized dataset for surrogate model training and testing. -Surrogate Model Design and Training: Develop a neural-network surrogate model capable of predicting the nonlinear response—including buckling behavior—from the microstructural parameters. Train, validate, and test the model to ensure high accuracy and generalization. -Parametric and Topology-Based Studies: Use the trained surrogate model to explore how variations in topology, geometry, or beam properties influence macroscopic structural performance. Identify trends in deformation and dissipated energy to provide physical insight into design–response relationships. -Comprehensive Documentation: Document the full research workflow—literature review, FEA modeling, dataset creation, surrogate model development, validation, and parametric analysis—in a structured and comprehensive final report.

Tema: ENGINYERIA AERONÀUTICA

Tipus: Estudi

Càrrega de treball:

The workload will depend on the Bachelor degree the student is enrolled in

Temàtica ambiental: No

Confidencial (informatiu): No

Ambit de cooperació: No

Publicació: 15/12/2025

Caducitat: 12/12/2026