Development of a Light Field Camera-Based Diagnostics Methodology for Fluid-Structure Interaction

Abstract

Traditional multi-camera FSI diagnostics are costly, complex to align, and limited by optical access, restricting their use in confined experimental settings for volumetric measurements. This dissertation introduces a single-camera methodology, termed light-field fluid-structure interaction (LF FSI), for simultaneous, time-resolved measurement of three-dimensional flow fields and structural deformation. Leveraging recent advances in high-speed light-field imaging, a single plenoptic system captures both fluid and structural phases simultaneously, and a POD-based image processing strategy decouples the combined information. This enables conventional plenoptic reconstruction methods to extract flow and structural information. The methodology is validated on synthetic datasets and controlled experiments with flexible structures, successfully capturing large-scale flow features, although depthwise velocity components remain less accurate. To further enhance reconstruction accuracy, a physics-informed neural network (PINN) framework is applied to regularize the flow field using governing equations and moving boundary conditions. In the flow around a flexible membrane, this approach enables recovery of out-of-plane velocity components and reduces noise across the flow field. The impact of boundary information is systematically evaluated using synthetic heart valve data, demonstrating the successful application of LF FSI in unsteady cardiovascular flows and the extraction of near-wall flow features. Finally, the methodology is applied to a cardiovascular flow environment in a heart valve simulator, capturing the coupled dynamics of pulsatile flow and leaflet motion and enabling characterization of large-scale flow structures. Overall, this work establishes a compact, experimentally feasible framework for volumetric FSI diagnostics using a single camera, enabling simultaneous measurements of flow structures in optically constrained environments.