Skip to content
Politecnico di Milano, SIGMA Lab
All people
Portrait of Alessandro Lucchetti

Alessandro Lucchetti

PhD Student · Cycle 40

Supervised by Francesco Cadini

alessandro.lucchetti@polimi.it
ORCID Google Scholar LinkedIn GitHub

Research

Alessandro's research develops graph neural networks as fast surrogates of finite element simulations in structural mechanics. His main line extends GNSS, a graph neural network simulator of guided wave propagation, towards structural health monitoring, with stable long-term predictions of elastic waves in structures made of different materials. A second line applies graph neural networks to the inverse design of disordered lattice metamaterials for protective structures: a surrogate trained on finite element crushing simulations, with the material calibrated on tests of the printed polymer, guides a generative model towards geometries with a prescribed crushing response. He also works on the Italian Space Agency project EXTESA.

Selected publications

  1. Towards Adaptive and Robust Unsupervised Anomaly Detection in Satellite Telemetry

    L. Brancato, A. Lucchetti, F. Cadini, M. Giglio

    PHM Society Asia-Pacific Conference 5(1) · 2026

  2. Fractal-based Satellite Health Monitoring

    L. Pinello, L. Brancato, A. Lucchetti, F. Cadini, M. Giglio

    PHM Society Asia-Pacific Conference 5(1) · 2026

Teaching

  • Teaching assistant

    Reliable and Resilient Design of Mechanical Systems

    MSc Mechanical Engineering · 5 CFU · 1st semester · Milano Bovisa

    Lecturer: Francesco Cadini

    Course sheet

  • Teaching assistant in 2025/2026

    Machine Design

    MSc Mechanical Engineering · 5 CFU · 2nd semester · Milano Bovisa

    Lecturer: Andrea Manes

    Course sheet

Courses of 2026/2027, from the Politecnico’s course sheets. All the group’s courses

Research in images

GNSS, a graph neural network that simulates structural dynamics: the structure is turned into a graph, and message passing between its nodes predicts the next state of the whole structure, one time step at a time.
GNSS, a graph neural network that simulates structural dynamics: the structure is turned into a graph, and message passing between its nodes predicts the next state of the whole structure, one time step at a time.
A guided wave travelling along a beam: the finite element simulation (top) and the GNSS prediction (bottom).
Disordered hexagonal lattices: the thickness of every strut is perturbed at random, from a uniform lattice (0 %) to 30 % disorder.
Disordered hexagonal lattices: the thickness of every strut is perturbed at random, from a uniform lattice (0 %) to 30 % disorder.
A 3D-printed TPU lattice with a disordered design crushed under quasi-static compression in the laboratory (twice the real speed).