Rockwell Adhesion Test - Approach to Standard Modernization

Abstract
Automatization of industry processes and analyses has been successfully applied in many different areas using varying methods. The basis for these industrial analyses is defined by global or country specific standards and often development of automated solutions works towards streamlining processes currently done heuristically. Lately, image classification, as one of the automatization development areas, has turned to machine learning in search for solutions. Though approaches that involve neural networks often result in high accuracy predictions, their complexity makes feature hard to understand and ultimately reproduce. To this end, we introduce a pipeline for the design, implementation and evaluation of a hand-crafted feature set used for the parameterization of two thin film coating adhesion classification standards. The method mimics the current expert classification process, and is developed in collaboration with domain experts. Implementation of an automated classification process was used for verification and integration testing.
Description

        
@inproceedings{
10.2312:eurp.20201121
, booktitle = {
EuroVis 2020 - Posters
}, editor = {
Byška, Jan and Jänicke, Stefan
}, title = {{
Rockwell Adhesion Test - Approach to Standard Modernization
}}, author = {
Hatic, Damjan
and
Cheng, Xiaoyin
and
Weibel, Thomas
and
Rauhut, Markus
and
Hagen, Hans
}, year = {
2020
}, publisher = {
The Eurographics Association
}, ISBN = {
978-3-03868-105-2
}, DOI = {
10.2312/eurp.20201121
} }
Citation