Structural health monitoring using dynamic mode decomposition
At LMSSC, Paris, February 12th 2025, 1.30 p.m.
Fernanda Thaís Colombo
Postdoctoral researcher, Mechanical Engineering Department (DEM), Structural Health Monitoring Laboratory (SHM Lab),
São Paulo State University (UNESP), Ilha Solteira, Brazil
Vision systems have been increasingly used in structural health monitoring because they allow non-contact and full-field vibration analysis compared to traditional sensors such as accelerometers and strain gauges. However, most image-based monitoring techniques rely on complex pre-processing of the images before extracting information about the system dynamics.
In this context, our research group at UNESP investigated applying dynamic mode decomposition (DMD), a data-driven and physics-free technique, directly in the image frames of a structure's motion to obtain the spatiotemporal patterns that can be observed in the data. These patterns are verified as coherent with the structure's dynamics and can be used to identify its current health condition.
Numerical and experimental tests were carried out using videos of a clampedfree beam under different integrity conditions. With good performance even for noisy images with low resolution and frame rate, this alternative allows for automatic health monitoring that is low-cost, non-contact, and with less processing complexity.
Biography:
Fernanda Thaís Colombo is a postdoctoral researcher at the Mechanical Engineering Department at São Paulo State University - UNESP (Ilha Solteira, Brasil), with a São Paulo Research Foundation (FAPESP) scholarship. She completed her bachelor, master and doctorate degrees in mechanical engineering at the São Carlos School of Engineering - EESC-USP (São Carlos, Brasil), specializing in Dynamics and Mechatronics.
She is currently doing a research internship at Arts et Métiers (Paris, France) with the goal of developing data-driven strategies to reduce vibration and mitigate the impact of damage to a structure. Her research interests include modeling of dynamical systems, control systems, system identification, signal processing, structural health monitoring, and computer vision.
More informations about Fernanda Thaís Colombo: