Deep-learning-based out-of distribution data detection in visual inspection images
Erik Lindgren, Christopher Zach
Antingen stödjer din webbläsare inte javascript, eller är javascript inaktiverat. Denna webbplats fungerar bäst om du aktiverar javascript.
Researcher
I am a researcher, postdoctoral fellow, within non-destructive testing (NDT) and evaluation (NDE). In short, it is about quality control of products without negatively affecting their function. This as a part of a production system, as part of lifetime extension of products already in-service, or as part of re-pair or re-use in circular economies.
My focus is mainly on X-ray-based methods, digital X-ray radiography in 2D and X-ray computed tomography (XCT) in 3D. In my postdoc I explore how modern artificial intelligence, e.g. deep learning, can be used industrially within OFP. This includes, among other things, questions about how an automated or semi-automated data interpretation can be performed in an industrially safe and efficient way, e.g. how to handle new unknown input far from the training data; but also, how mathematical modeling of the inspection can be used to create added value within OFP, e.g. calculation of capability and measurement uncertainty.
The applications have a focus on additively manufactured metal products and welds. In most of my projects, I collaborate with the industry.
I teach in courses on nondestructive testing and evaluation, in our NDE course but also with NDE parts in other manufacturing courses at the University.
NDT, NDE, X-ray computed tomography, digital industrial X-ray, artificial intelligence.
Publications
RESEARCH PROJECTS