Faults with the internal elements of rotating equipment are notoriously difficult to diagnose while the equipment is in service, and so maintainers often have to rely on indirect assessment based on changes in performance and vibrations. A novel approach to visualizing faults with rotating components is presented based on penetrating electromagnetic radiation and image construction. The basic concept of operations is to generate photons from a collimated source that is relocatable so as to illuminate the rotating component in a piecewise fashion along its radius, preferably but not necessarily parallel to the axis of rotation. Subimage acquisition captures stationary parts through which photons pass as well as part of the rotating component. Registration of rotating sub-images is then done as the machine rotates at quasi-steady-state speed to construct the complete image. Widefield high-resolution Fourier ptychographic tomography is employed to stitch together variably illumined low-resolution sub-images in Fourier space to improve resolution and reduce vignetting effects and other artifacts (such as shadowing due to the stationary components). Image acquisition rate and source power can be adjusted to reduce smearing of the parts that are rotating. From the final image, artifact feature extraction is then used to create fault primitives from which fault detection and identification can be done. A proof-of-concept system is presented with experimental characterization using idealized geometries (black line segments and polygons, gray-scale images). A path to practical implementation using x-ray sources and robotic detector positioning is described.
Bauer et al. (Wed,) studied this question.