RT info:eu-repo/semantics/article
T1 Data augmentation in predictive maintenance applicable to hydrogen combustion engines: a review
A1 Schwarz, Alexander
A1 Rodríguez Rahal, Jhonny
A1 Sahelices Fernández, Benjamín
A1 Barroso García, Verónica
A1 Weis, Ronny
A1 Duque Antón, Simón
K1 Data augmentation
K1 Predictive maintenance
K1 Anomaly detection
K1 Generative models
K1 33 Ciencias Tecnológicas
AB Machine-learning-based predictive maintenance models, i.e. models that predict break-downs of machines based on condition information, have a high potential to minimizemaintenance costs in industrial applications by determining the best possible time to per-form maintenance. Modern machines have sensors that can collect all relevant data of theoperating condition and for legacy machines which are still widely used in the industry,retrofit sensors are readily, easily and inexpensively available. With the help of this datait is possible to train such a predictive maintenance model. The main problem is thatmost data is obtained from normal operating conditions, whereas only limited data arefrom failures. This leads to highly unbalanced data sets, which makes it very difficult,if not impossible, to train a predictive maintenance model that can detect faults reliablyand timely. Another issue is the lack of available real data due to privacy concerns. Toaddress these problems, a suitable data generation strategy is needed. In this work, a litera-ture review is conducted to identify a solution approach for a suitable data augmentationstrategy that can be applied to our specific use case of hydrogen combustion engines inthe automotive field. This literature review shows that, among the different state-of-the-artproposals, the most promising for the generation of reliable synthetic data are the onesbased on generative models. The analysis of the different metrics used in the state of theart allows to identify the most suitable ones to evaluate the quality of generated signals.Finally, an open problem in research in this area is identified and it is the need to validatethe plausibility of the data generated. The generation of results in this area will contributedecisively to the development of predictive maintenance models.
PB Springer
YR 2024
FD 2024
LK https://uvadoc.uva.es/handle/10324/75222
UL https://uvadoc.uva.es/handle/10324/75222
LA eng
NO Artificial Intelligence Review, 2024, vol. 58, n.1
NO Producción Científica
DS UVaDOC
RD 12-mar-2025