Abstract:
Maintenance is a fundamental activity performed on machines and industrial equipment to
optimize the utilization of resources. Over the years, maintenance techniques such as corrective,
preventive, and predictive maintenance have emerged with industrial revolutions. One crucial
aspect in maintenance is assessing the remaining useful life of assets. Remaining useful life refers
to the estimated time or usage that a product or asset could continue to operate effectively and
efficiently before becoming obsolete or requiring replacement. Predictive maintenance is an
advanced maintenance approach that comprehended the concepts of the fourth industrial
revolution to accurately compute the equipment residual life. Latest emerging technologies such
as cyber-physical systems, internet of things, big data, and smart production played a key role in
the practical implementation of predictive maintenance, particularly in the manufacturing sector.
Therefore, there is a need to continuously improve data-driven deep learning algorithms to
accurately predict the remaining useful life. This research thesis presents an improved deep
learning model for the remaining useful life prediction of a transportation system using long-short
term memory network and hyperparameter tuning through Bayesian optimization. The findings of
the research indicated lowest root mean square error due to improved preprocessing and
hyperparameter tuning. It significantly improved equipment reliability, reduced downtime, and
lowered maintenance costs by leveraging data-driven insights to predict and prevent failures before
they occur. Moreover, predictive maintenance has some issues that need to be resolved for its full
practical implementation. For this purpose, the challenges and future opportunities are also listed
at the end
