Universidade Federal do Ceará
Mestrado e Doutorado em Ciências da Computação

Título: An Enhanced Hard Drive Lifetime Prediction Based on an Anomaly Detection Framework

Data: 13/02/2025
Horário: 16h
Local: Sala de Seminários – Bloco 952

Resumo:

Remaining Useful Life (RUL) prediction for a Hard Disk Drive (HDD) is a challenging, yet important task to ensure the service level of data storage providers. In recent years, many authors proposed machine learning-based methods to address this problem. However, by analyzing previous works, one can notice that the degradation state of each HDD is usually unknown and such an aspect is not considered in the proposed methods. In this work, we present a RUL prediction framework that initially detects anomalies, that may indicate an incipient degradation, and then predicts the RUL. Our framework is used to select training instances and predict the remaining useful life of test HDDs using Recurrent Neural Networks. In addition, we propose the clustering of the selected data instances to create local models based on how the HDD data degrades. Based on our experiments, we verify that the proposed framework led to significant improvements in RUL prediction metrics.

Banca examinadora:

• Prof. Dr. Javam de Castro Machado (MDCC/UFC) – Orientador
• Prof. Dr. João Paulo Pordeus Gomes (UFC) – Coorientador
• Prof. Dr. César Lincoln Cavalcante Mattos (MDCC/UFC)
• Prof. Dr. João Paulo do Vale Madeiro (MDCC/UFC)