CFD-Guided Restoration of Occluded Schlieren Images for Supersonic Turbine Cascade Flow Diagnostics

The operating environment of large-scale steam turbines is complex, and the flow characteristics around the long blades of the final stage are critical to both plant efficiency and safety. However, the fixed structure of ultra-thin blades causes occlusions in shadowgraph images, severely limiting precise analysis of flow characteristics. To this end, this paper proposes a physics-informed restoration framework specifically designed to overcome large-scale physical occlusions in ultra-thin blade cascade experiments. Distinct from general inpainting, our method achieves strict spatial coordinate synchronization between experimental and numerical domains via affine mapping, ensuring the reconstructed flow features are anchored to a consistent physical coordinate system. Combined with edge detection and feature recognition, occluded regions are localised, and numerical Schlieren information is utilised to reconstruct and repair these areas. Research findings demonstrate that this method effectively restores obscured regions, significantly improves the quality of the Schlieren image and provides reliable data support for in-depth analysis of flow characteristics of the final long blade cascade of steam turbines.