-
Leonardo Souza Coelho
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Universidade Federal De Itajubá - UNIFEI - Itabira
-
Giovani Bernardes Vitor
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Universidade Federal De Itajubá - UNIFEI - Itabira
-
Willian Gomes de Almeida
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Universidade Federal De Itajubá - UNIFEI - Itabira
-
Rafael Francisco dos Santos
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Universidade Federal De Itajubá - UNIFEI - Itabira
-
Paulo Henrique Vieira Soares
Gerência de Serviço de Tecnologia de Automação Sudeste, VALE S/A; Itabira, Minas Gerais
Keywords:
Ore Drying, Artificial Intelligence, Computer Vision, Automation 4.0, Mining
Abstract
Industry 4.0 has provided significant advances in several areas, one of them being in the mining sector. Within the mining process, the ore drying stage is characterized as an important phase in which the ore passes through rotating vacuum filters, in order to reduce its moisture. One of the problems at this stage concerns the detection of the quality of filters in the drying machine, with the clogging of the filtering medium being the main reason for the poor quality of the filter. Therefore, this work proposes the use of deep neural networks to recognize the quality of these filters for preventive maintenance. Unet and SegNet network architectures were applied, two types of convolutional neural networks, encoder/decoder. As a result, it was possible to obtain more than 90% accuracy in the correctness of the model, as well as the determination of a quality indicator for these equipments.
