Complex Decay Prediction of Marine Machinery Using Multilabel SVM

Abstract

In this article, a multilabel support vector machine (SVM)-based approach is investigated to address the simultaneous decay detection of the marine propulsion system. To verify the performance of the algorithm, we perform some experiments using a simulation dataset from a real-data validated numerical simulator of a Frigate. In particular, we try to train the model without simultaneous decay data, considering the great difficulty of obtaining simultaneous decay data in practice. The experimental results show that the proposed approach can identify the complex decay modes of the marine propulsion system effectively using only simple decay data in the training process. Introduction The propulsion system is considered to be the “heart” of a marine ship (Li et al. 2019a). Its safety and reliability are critical to the regular operation of the ship (Bayer et al. 2018; Cheliotis & Lazakis, 2018; Lazakis et al. 2016). However, performance decay may occur to the propulsion system due to the high humidity and high salt characteristics of the marine environment (Fang et al. 2018; Kang et al. 2019; Wang et al. 2019). The decay modes can be divided into single decay and simultaneous decay. Single decay indicates a simple decay mode that only one kind of decay occurs at a time, and simultaneous decay indicates a complex decay mode that multiple decays occur at the same time. To improve the safety and reliability of the marine propulsion system, researchers have proposed many related approaches from the perspective of fault diagnosis.