Agile Surface Inspection Framework for Aerospace Components Using Unsupervised Machine Learning

Abstract

Quality Control is an important step in manufacturing machined parts, especially complex, customized parts for safety critical systems such as airplane engines. Visual Inspection by humans is one of the modalities used for this purpose, but it is limited in many ways. Inspectors need months of training, need to maintain tremendous focus for a long duration, and are required to keep up with the growing pace of manufacturing. It is thus imperative to automate this process. This article proposes a flexible automated path planning framework which can be adapted to any robot, which implements a mesh segmentation algorithm and attempts to generalize the custom solution. The paper describes elements of the solution, and evaluates its efficacy pertaining to geometries of parts with characteristics similar to components found in aerospace. Furthermore, this article explores a potential improvement to the automated inspection process.

Publication
Proceedings of ASME/ISCIE International Symposium on Flexible Automation