Title: The use of neural networks for the classification of casting defect

Authors: Leszek A. Dobrzanski, Mariusz Krupinski, Jerry Sokolowski, Piotr Zarychta

Addresses: Silesian University of Technology, Konarskiego Street 18a, Gliwice 44-100, Poland. ' Silesian University of Technology, Konarskiego Street 18a, Gliwice 44-100, Poland. ' University of Windsor Industrial Research Chair in Light Metals Casting Technology, 401 Sunset Avenue, Windsor, Ontario N9B 3P4, Canada. ' Silesian University of Technology, Konarskiego Street 18a, Gliwice 44-100, Poland

Abstract: Employment of the artificial intelligence tools for development of the methodology of assessing quality and structural defects in Al and Mg alloys and custom-made computer software will make it possible to determine the quality of manufactured element based on digital images registered in the X-ray flaw detection examinations. The possibility to correlate the frequency and morphology of defects with the technological process parameters will make it also possible to identify and classify these defects and control the process to minimise and eliminate them. The methodology is presented in this paper, making it possible to determine the types and classes of defects developed during casting elements from aluminium alloys, making use of photos obtained from the flaw detection method with X-ray radiation. [Received 10 November 2005; Accepted 9 January 2007]

Keywords: research methodology; Al-Si-Cu; neural networks; image analysis; casting defects; defect classification; X-ray flaw detection; aluminium alloys; process parameters.

DOI: 10.1504/IJCMSSE.2007.013845

International Journal of Computational Materials Science and Surface Engineering, 2007 Vol.1 No.1, pp.18 - 27

Published online: 30 May 2007 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article