A one-class SVM based approach for condition-based maintenance of a naval propulsion plant with limited labeled data

Maintenance of the main propulsion system is an important part of ship operation, whose costs account for the most of the costs of ship machinery maintenance. Related literature has shown that condition-based maintenance can minimize maintenance costs and machinery downtime. In this respect, we propose a one-class support vector machine based approach for machinery decay status assessment which helps to achieve condition-based maintenance. The novelty of our work lies in the using of decision values returned by the trained models to estimate the decay degree and the main decay direction. In particular, only the normal data and a small amount of labeled decayed data is necessary in model training process which greatly reduces the requirement for labeled data. Consequently, it can be used as a decision support tool for condition-based maintenance not only on this plant, but also on many other machinery.