Enhancing Remaining Useful Life (RUL) Prediction through CNN-Based Fusion Architectures

Phyo, Zin Win (2025) Enhancing Remaining Useful Life (RUL) Prediction through CNN-Based Fusion Architectures. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (760kB) | Preview

Abstract

Predictive Maintenance (PdM) enables proactive detection of equipment degradation and reduction of downtime and operation costs. Remaining Useful Life (RUL) estimation is central to PdM using sensor-derived time-series data to predict the time until an equipment fails. Although Convolutional Neural Networks (CNNs) are capable of learning the spatial features effectively, their limited capacity to capture long-term temporal dependencies can restrict the prediction accuracy. This study evaluated three CNN-based fusion architecture, which are CNN+Transformer Encoder, Multi-Scale CNN, and CNN+ Autoencoder on the NASA CMAPSS dataset. This was done using a unified experimental framework to ensure consistent preprocessing, training, and evaluation. Model performance was evaluated in terms of predictive accuracy, computational efficiency, and model complexity. Results indicate that every fusion model outperform the baseline standalone CNN model, with CNN + Autoencoder delivering the highest accuracy, Multi-Scale CNN performing a good trade-off between performance and efficiency, and CNN + Transformer enhancing temporal modelling capabilities at the expense of greater computational demand. These findings provide practical guidance for selecting CNN-based fusion strategies to implement the PdM system on the real-world industrial deployment, balancing predictive capability with resource requirements.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Niculescu, Hamilton UNSPECIFIED
Uncontrolled Keywords:	Predictive Maintenance (PdM); Remaining Useful Life (RUL); Convolutional Neural Networks; Transformer Encoder; Multi-Scale CNN; Autoencoder
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Telecommunications > Computer networks > Internet of things
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	02 Jul 2026 14:48
Last Modified:	02 Jul 2026 14:48
URI:	https://norma.ncirl.ie/id/eprint/9447

Actions (login required)

View Item