ABSTRACT

The steel industry presents several problems and opportunities for improvement, from the factory floor to the business management level. Operational procedures are continually improved to reduce failures, create reliable parameters and increase the reliability of the equipment. A highlight is the computer vision, presented in several processes, contributing to the continuous and accelerated advancements of innovations in industrial processes, allowing systems' automation or upgrade and changing their way of operation. This project aims to segment and detect, through convolutional neural networks, the wear of the shovels of the slag scrapers in pig iron pans in a Kambara Reactor of an industrial steel plant. In other words, the goal is to detect the wear of the shovels to control their use and replacement using mask R-CNN (Regionbased Convolutional Neural Network), for instance, segmentation and pixel count for wear control and change forecast.

摘要

钢铁行业从工厂车间到企业管理层面都存在一些问题和改进的机会。操作程序不断改进，以减少故障、创建可靠的参数并提高设备的可靠性。一个亮点是计算机视觉，它体现在多个​​流程中，有助于工业流程创新的持续和加速进步，允许系统自动化或升级并改变其操作方式。该项目旨在通过卷积神经网络分割和检测工业钢厂神原反应堆生铁锅中刮渣机铲子的磨损情况。换句话说，