| KEYWORD |
Physics-Informed Machine Learning Models for Shape Sensing
Reference persons CECILIA SURACE
Thesis type NUMERICAL
Description Shape sensing, the process of determining the geometry of a deformable structure from a set of sparse measurements, is a critical challenge in fields like aerospace, and structural health monitoring. Traditional methods often rely on complex models and computationally expensive numerical solvers, making them unsuitable for real-time applications. This thesis proposes a novel approach to the inverse shape-sensing problem using Physics-Informed Neural Networks (PINNs). Unlike conventional neural networks, PINNs incorporate the underlying physical laws directly into the training process, leveraging partial differential equations (PDEs) that govern the deformation behaviour of the structure. By embedding this domain knowledge, the proposed PINN model can infer the full-field shape from sparse and noisy measurement data while respecting physical constraints, thereby enhancing prediction accuracy and generalizability. This work will focus on developing a PINN-based framework that can solve inverse shape-sensing problems across various applications. The research aims to validate the model through a series of simulated and experimental case studies, demonstrating its capability to achieve high accuracy and computational efficiency.
Deadline 18/11/2025
PROPONI LA TUA CANDIDATURA