Logo do repositório

Percorrer por autor "Duarte, Miguel"

A mostrar 1 - 2 de 2
  • Enhancing predictive accuracy in aircraft engine mro using clustering and similarity methods
    Publication . Mendonça, Leonardo; Pires, Flávia; Barbosa, José; Duarte, Miguel; Leitão, Paulo
    In the aircraft engine Maintenance, Repair, and Overhaul (MRO) process, effective task planning relies heavily on the expertise of lead engineers. However, when predictive models are used to assist decision-making, issues with incomplete, unbalanced, and inconsistent data can lead to errors in the planning task. Therefore, reliable predictions are crucial for optimising the operational efficiency. This paper proposes a methodology to improve the prediction of maintenance times for an aircraft engine MRO process by integrating K-means clustering with Cosine and Jaccard similarity methods to define a reliable prediction interval. This methodology was compared with the predictions of the Simple Linear Regression method, which resulted in the prediction interval approach significantly reducing prediction errors, increasing prediction accuracy, while optimising process management and maintenance task planning throughout the MRO process.
    2025Documento de conferência Acesso aberto Ver mais
  • Monitoring and prediction of maintenance operations for aircraft engines repair
    Publication . Mendonça, Leonardo; Pires, Flávia; Duarte, Miguel; Barbosa, José; Leitão, Paulo
    Accurately estimating the hours required for maintenance, repair and overhaul (MRO) operations in the aviation sector frequently depends on the experience and personal judgment of engineers, can lead to introducing errors, increased operating costs, and time-consuming decision-making. This work presents the development of a cost-effective application to monitor and predict MRO operations in an aeronautical company. The application integrates data-driven algorithms, particularly Machine Learning (ML), with Power BI to provide a dynamic and user-friendly visualisation of historical and predicted data, improving decisionmaking time and facilitating operational planning. The simple linear regression model was the most effective algorithm to predict MRO operation for the case study with a B? of 0.81, balancing simplicity and performance compared to other analysed models.
    2025Documento de conferência Acesso aberto Ver mais