Speaker
Description
Autonomous ships are anticipated to support the shipping sector's sustainability and resilience. However, the health monitoring and maintenance management barriers for autonomous ship machinery must be sufficiently addressed. Prognostics and health management (PHM) approaches to assess the health conditions of autonomous ship machinery can provide solutions for these barriers. However, the lack of appropriate datasets representing a wide envelope is required to develop PHM models. Furthermore, the trained PHM model may not work appropriately due to the different operating conditions between the development stage and the application stage. This study aims to develop an intelligent health assessment system for autonomous ship machinery systems employing a marine dual-fuel engine. This is carried out by employing a simulation-based data generation method using a digital twin of high fidelity to derive the required datasets for the engine. The engine operations at both healthy and faulty conditions were considered including the anomalies of the intake and exhaust valve leakages. Additionally, the engine ambient conditions and engine operating loads were approximated by analysing available historical records. The trustworthiness of the digital twin and generated datasets is addressed by a novel framework including validation, verification, and robustness. The PHM system consists of the diagnosis and prognosis data-driven models. The former is based on a support vector machine (SVM) and its function is to detect, identify, and isolate valve leakage faults. The latter employs the deep neural network (DNN) to estimate and predict a health indicator for the valves’ degradations. The developed anomaly diagnosis and prognosis models are trained and tested employing simulation-generated datasets exhibiting accuracy of over 90% based on the R-squared metric. Additionally, the developed PHM models are applied to validation datasets representing the extended operating envelope for investigating the validity of the PHM model in practical operations. This study provides insights for the development of future intelligent decision-making systems for autonomous ship machinery to support their health-aware and fault-tolerant system management.
References
-
| Keywords | Prognostics and Health Management, Digital Twins |
|---|
