Extracting Fluid--Structure Modes in Turbulence-Driven FIV From Particle Tracking and Shape Sensing Data

Flow-induced vibrations (FIVs) arise in engineering and biological systems ranging from offshore risers to arterial vessels. They are challenging to characterize due to strong fluid--structure coupling and sparse, often single-phase measurements. We investigate a model FIV scenario, with a flexible strut immersed in the turbulent wake of a cylinder within a free-surface flume. Structural motion is captured using kinematic shape sensing (KSS) beams, while fluid measurements are obtained via particle tracking velocimetry (PTV). A physics-informed neural network (PINN) is trained to reconstruct the 3D velocity and pressure fields by minimizing residuals from the Navier--Stokes equations, a moving-wall boundary condition derived from KSS data, and a particle tracking loss that refines the Lagrangian data. This optimization produces physically consistent reconstructions that match the measured particle trajectories and beam deflections. The resulting flow fields will be analyzed using spectral proper orthogonal decomposition and stochastic subspace identification, enabling identification of synchronous fluid--structure modes. We present initial results from this PINN--co-analysis framework and discuss implications for realistic, turbulence-driven FIV.