This paper describes a novel methodology concerning the application of machine learning for the intelligent monitoring of ship systems. Monitoring of machinery condition is a crucial aspect of maintenance optimisation that is required for the vessel operation to remain sustainable and profitable. Operator-approved performance data are used to train a data-driven Expected Behaviour Model (EBM). Once trained, this model accepts newly-acquired data points as input and returns the probability of them belonging to the same performance profile with training data. Through this, emerging anomalies can be detected. This tool is coupled with a short-term healthiness forecasting tool, able to estimate future healthiness index. This combination allows the derivation of a healthiness score of both current and future condition of ship machinery. Additionally, in cases of performance degradation, the Remaining Useful Life (RUL) of given system can be projected. This provides a robust framework for the early detection of incipient machinery faults.
|Title of host publication||Proceedings of the 2018 Smart Ship Technology Conference|
|Place of Publication||London|
|Publisher||Royal Institution of Naval Architects|
|Number of pages||6|
|Publication status||Published - 23 Jan 2018|
|Event||Smart Ship Technology 2018 - London, United Kingdom|
Duration: 23 Jan 2018 → 24 Jan 2018
|Conference||Smart Ship Technology 2018|
|Period||23/01/18 → 24/01/18|
- machine learning
- remaining useful life
- expected behaviour model
Gkerekos, C., Lazakis, I., & Theotokatos, G. (2018). Exploiting machine learning for ship systems anomaly detection and healthiness forecasting. In Proceedings of the 2018 Smart Ship Technology Conference London: Royal Institution of Naval Architects.