Machine Learning Applied to Industrial Machines for an Efficient Maintenance Strategy: A Predictive Maintenance Approach

Maintenance activities are crucial in manufacturing environments to reduce machine breakdowns and maintain product quality. However, traditional maintenance strategies can be expensive, as they can lead to unnecessary maintenance activities. As a result, Predictive Maintenance (PdM) can be a great way to solve these issues, as it enables the prediction of a machine's condition/lifespan allowing for maintenance-effective manufacturing. This paper aims to address these issues by proposing a novel methodology to improve the performance of PdM systems, by proposing a machine learning training methodology, an automatic hyperparameter optimizer, and a retraining strategy for real-time application. To validate the proposed methodology a random forest and an artificial neural network model are implemented as well as explored. A synthetic dataset, that replicates industrial machine data, was used to show the robustness of the proposed methodology. Obtained results are promising as the implemented models can accomplish up to 0.97 recall and 93.15% accuracy.