Marcelo Aguiar, Marley Vellasco, Ricardo Tanscheit, Marco Pacheco, and Manoela Kohler, Pontifical Catholic University of Rio de Janeiro, R. Marquˆes de S˜ao Vicente, 225, 22451-900 Rio de Janeiro/RJ, Brazil
Since the beginning of the 21st century, predictive maintenance (PdM) has gained increasing prominence in the industry, providing the capability to predict the Remaining Useful Life (RUL) of equipment and mitigate accidents and financial losses. In this context, Machine Learning (ML) models are widely employed. This study proposes the prediction of Turbofan engine RUL through a machine learning model trained with historical data from this equipment. Two distinct models were evaluated: the Fuzzy System and the Neuro-Fuzzy System. To enable training, it was necessary to select the best features using a genetic algorithm, aiming to reduce complexity and enhance the performance of the Fuzzy System. The model with the best performance was the Fuzzy System with 5 selected features and 11 Fuzzy sets. Despite not yielding the lowest RMSE metric compared to related works, this interpretable model demonstrated reasonable performance in comparison to the model by Babu et al. [1]. This suggests that, for the dataset in question, the Fuzzy system is recommended to ensure better interpretability, while neural networks used in previous studies are more suitable for more precise predictions.
Fuzzy System, Genetic Algorithm, Machine Learning, Turbofan, Predictive Mainte-nance.
Quentin Jayet1, Christine Hennebert1, Yann Kieffer2, and Vincent Beroulle2, 1Univ. Grenoble Alpes, CEA, LETI, DSYS, 38000 Grenoble, France, 2Univ. Grenoble Alpes, Grenoble INP, LCIS, 26000 Valence, France
Consensus protocols and peer-provided proofs establish trust in blockchain systems. This paper investigates the role of the elapsed time, i.e., the guarantee that one peer on the network can propose a block within a predefined time interval, on creating trust-by-design. Unlike energyintensive Proof of Work (PoW) used in Bitcoin, verifiable proofs based on security components have emerged as an alternative. However, PoW enables parallelisation and delegation, which can lead to recentralisation of the protocol. This paper introduces Proof of Hardware Time (PoHT), executed on a System on Module (SoM) with an ARM Cortex-A7 and an external Trusted Platform Module (TPM). Additionally, this paper compares PoHT, PoW, and Proof of Sequential Work (PoSW) in a Trusted Execution Environment (TEE) based on their elapsed time distribution and energy consumption. Finally, PoHT is compared with existing proof mechanisms.
Blockchain, Internet of Things, Elapsed time, Trusted Execution Environment, Verifiable proof, Energy consumption, Trusted Platform Mod.
Copyright © BIoT 2024