Dokument-ID Dokumenttyp Verfasser/Autoren Herausgeber Haupttitel Abstract Auflage Verlagsort Verlag Erscheinungsjahr Seitenzahl Schriftenreihe Titel Schriftenreihe Bandzahl ISBN Quelle der Hochschulschrift Konferenzname Quelle:Titel Quelle:Jahrgang Quelle:Heftnummer Quelle:Erste Seite Quelle:Letzte Seite URN DOI Abteilungen OPUS4-1852 Wissenschaftlicher Artikel Boersch, Ingo; Füssel, Uwe; Gresch, Christoph; Großmann, Christoph; Hoffmann, Benjamin London, Springer Data mining in resistance spot welding - A non-destructive method to predict the welding spot diameter by monitoring process parameters Resistance spot welding is the dominant process in the present mass production of steel constructions without sealing requirements with single sheet thicknesses up to 3 mm. Two of the main applications of resistance spot welding are the automobile and the railway vehicle manufacturing industry. The majority of these connections has safety-related character and therefore they must not fall below a certain weld diameter. Since resistance spot welding has been established, this weld diameter has been usually used as the gold standard. Despite intensive efforts, there has not been found yet a reliable method to detect this connection quality non-destructively. Considerable amounts of money and steel sheets are wasted on making sure that the process does not result in faulty joints. The indication of the weld diameter by in-process monitoring in a reliable way would allow the quality documentation of joints during the welding process and additionally lead through demand-actuated milling cycles to a substantial decrease of electrode consumption. An annual, estimated reduction in the seven- to nine-figure range could be achieved. It has an important impact, because the economics of the process is essentially characterized by the electrode caps (Klages 24). We propose a simple and straightforward approach using data mining techniques to accurately predict the weld diameter from recorded data during the welding process. In this paper, we describe the methods used during data preprocessing and segmentation, feature extraction and selection, and model creation and validation. We achieve promising results during an analysis of more than 3000 classified welds using a model tree as a predictor with a success rate of 93 %. In the future, we hope to validate our model with unseen welding data and implement it in a real world application. 2016 The International Journal of Advanced Manufacturing Technology 10.1007/s00170-016-9847-y Fachbereich Informatik und Medien