Review of the safety engineering techniques for a complex ship system

Abstract

Marine industry is leaning towards the autonomous vessels; and advanced technologies are being developed for autonomous operations. However, this rapid technological change has increased the level of complexity in ship systems. As the interactions between components are increasing further and software are getting imbedded into components, the nature of risks in modern systems can be different than in the traditional systems; where the risks were mostly limited to human errors and component failures. However, for identifying risks in modern systems, it is first important to understand the system composition and the behavior of components. Since traditional system-safety engineering techniques, developed for the relatively simpler systems in past, are still dominant in marine industry. These techniques may not be able to cope with the risks due to increasing complexity.This paper reviews and identifies a suitable modelling approach and a risk analysis method for a complex ship system. A modern modeling approach known as Systems-Modeling Language (SysML) and a modern risk analysis method known as Systems-Theoretical Process Analysis (STPA) are reviewed and compared with widely used traditional methods known as the Tree structure method and Fault Tree Analysis. SysML is a graphical modeling language that presents structural composition, component functions, behavior, constraints and requirements of a complex system. STPA is a risk analysis method that aims to identify and mitigate risks in a complex system. The review and comparison results are presented in the paper.The results of this study suggest that the modern methods are more suitable than the traditional methods when the functionality of each method are considered. However, as the modern methods are more detailed, and are focused on the functionality, they are relatively complex and require more resources for the analysis in comparison to the traditional methods. Some viable solutions to improve the drawbacks of SysML and STPA, and possible future research topics are presented.