Maintenance Optimization Through predictive Maintenance: A Case Study For Damen Shipyards

We use predictive modelling techniques to predict the remaining useful life of critical components on ships. Specifically, we use a recurrent neural network and a gradient boosting tree to predict the remaining useful life of the fuel separator. We found the gradient boosting tree to be a suitable model for this prediction task. When optimizing a maintenance threshold, the boosting tree outperforms a baseline time-based preventive maintenance model (a 1.4% cost reduction).