ABSTRACT
With Industry 4.0, the integration between physical and digital
environments enabled some improvements within several production
segments. Among these improvements, advancements on the
application of machine learning algorithms to predict current and
future states of equipment have been gaining attention, specially,
for maintenance purposes. This research work presents a comparative
experimental study on machine learning algorithms applied
to classification of industrial machinery states. After training and
evaluating models based on five different algorithms (i.e., Decision
Tree, Naive Bayes, Support Vector Machines, XGBoost and Neural
Network), some interesting results were obtained. Considering the
accuracy, precision, recall and training time of each model, it was
observed that some models performed well, while others may not
be as suitable for solving the problem. Such good performing models
could be used to schedule interventions on a given industrial
equipment, avoiding production stoppages.