Data mining for burr detection (in the drilling process)

Abstract

Drilling is the most important operation in aeronautic industry carried out previous to riveting. Its main problem lies with the burrs. Nowadays, there is a burr elimination task (manual task) subsequent to drilling and previous to riveting that increases manufacturing cost. It is necessary to develop a monitoring system to detect automatically and on-line when the generated burr is out of aeronautic limits, and then deburring. This system would reduce holes deburring to the holes which really are out of tolerance limits, focusing in trying to avoid false negatives. The article shows an improvement in burr generation prediction, using Data Mining techniques versus current mathematical model. It gives an overview of the process from data preparation and selection to data analysis (with machine learning algorithms) and evaluation of the models. © 2009 Springer Berlin Heidelberg.