Abstract: This paper presents a case study of application data mining techni-ques in synthetic data for predictive maintenance of a naval propulsion system,with the objective of analyzing its applicability and suitability in the construc-tion of predictive models for maintenance. In the first stage, we applied datamining techniques to the original dataset, and raised hypotheses about the re-sults obtained with synthetic data. In the second and third stages, respectively,we tested the hypotheses raised in the initial stage by inserting class imbalanceand measurement uncertainties. This way, we made the synthetic data moreuseful for building failure predicting models for real industrial scenarios.

Keywords: Predictive maintenance, Synthetic data, Data mining

Authors: Rafael Schena, João Cesar Netto and Karin Becker

Journal/Conference: Brazilian Symposium on Databases (SBBD)

Publication: https://doi.org/10.5753/sbbd.2022.224627