Supervised learning classification for dross prediction in ductile iron casting production
Foundry is one of the key axes in society because it provides with important pieces to other industries. However, several defects may appear in castings. In particular, Dross is defect that is a type of non-metallic, elongated and filamentary inclusion. Unfortunately, the methods to detect Dross have to be performed once the production has already finished using quality controls that incur in a subsequent cost increment. Given this context, we propose the first machine-learning-based method able to foresee Dross in iron castings, modelling the foundry production parameters as input. Our results have shown that this method obtains good accuracy results when tested with real data from a heavy-section casting foundry.
Autores/as:
Igor Santos (University of Deusto), Javier Nieves (AZTERLAN), Pablo G. Bringas (University of Deusto), Argoitz Zabala (AZTERLAN), Jon Sertucha (AZTERLAN)
Keywords:
Fundición, grafito, hierro, Kernel