Detection and segmentation of pig iron slag scrapers using Mask R- CNN for wear control

Resumo

The steel industry presents a vast list of problems and opportunities for improvement, ranging from the
factory floor to levels of business management. Operating procedures are revolutionized every day to decrease
failures, create reliable parameters and increase equipment reliability, and with the continuous and accelerated
advance of innovations in industrial processes, computer vision is increasingly present and is necessary for the
automation of new systems or of systems that need an update in their way of operating. This project aims to
segment and detect, through convolutive neural networks, the shovels of the slag scrapers in pig iron pans in a
Kambara Reactor of a steel plant. Aiming at detecting the wear of the shovels to control their use and replacement
using Mask R-CNN for instance segmentation and pixel count for wear control.