Construção de script de predição de falhas em ferramentas de usinagem utilizando o algoritmo de aprendizagem de máquina k-vizinhos-mais próximos

Abstract

This paper presents the development of a maintenance prediction algorithm for failures in machining tools with the aim of avoiding process stops due to unforeseen breakdowns, improving maintenance routines, and reducing corrective maintenance costs. The system was developed using concepts of artificial intelligence, business intelligence, predictive maintenance, and machine learning. The project was developed using the R language to execute the algorithm, and the data used were based on the database of the Machine Learning and Intelligent Systems Learning Center at the University of California. To predict failures, supervised K-Nearest-Neighbors learning models were used. The obtained results are presented.