Original title: Accurate Reconstruction of Gas Turbine Blade Geometry Using 3D/2D Rigid Registration and CT View Optimization
Hristo Valtchanov, Nicolas Piché, Vladimir Brailovski, Justin Byers +2 · September 2026
Gas turbine blades have intricate internal cooling channels that protect them from extreme heat. Inspecting these hidden shapes usually requires full CT scans, which take many images. Researchers developed a faster alternative. Instead of building a full 3D scan, they match a 3D digital model of the blade directly onto a few 2D X-ray pictures. Their method uses a stepwise algorithm that first lines up the outer shell and then aligns the inside pieces. It works by adjusting the model until simulated X-rays match the real images as closely as possible. They found that taking just a small number of tilted X-ray views gives the highest accuracy. The process reached errors under one-fifth of a pixel pitch and handled noise better than full CT scans.
Imagine you have a model car and you're trying to match it to its shadow on a wall, but the shadow is faint and distorted. Normally, you'd look from one spot and guess, but the distortion makes it hard to tell if parts are in the right place. In this paper, the researchers use multiple X-ray views like multiple shadows from different lights, and they align the model to these shadows step by step, starting with the outside, to get a precise fit even with noise.
Manufacturers can inspect parts like turbine blades without costly CT scans, saving time and money. This directly impacts the safety and efficiency of engines used in aviation and power generation.
The method is reduced in precision by image noise and defects, though it mitigates these effects better than CT reconstruction, and it is presented as an alternative, not a full replacement for all inspection scenarios.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Accurate Reconstruction of Gas Turbine Blade Geometry Using 3D/2D Rigid Registration and CT View Optimization — Hristo Valtchanov, Nicolas Piché, Vladimir Brailovski et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.