Machine Learning and Case-Based Reasoning for Real-Time Onboard Prediction of the Survivability of Ships

Abstract

The subject of damaged stability has greatly profited from the development of new tools and techniques in recent history. Specifically, the increased computational power and the probabilistic approach have transformed the subject, increasing accuracy and fidelity, hence allowing for a universal application and the inclusion of the most probable scenarios. Currently, all ships are evaluated for their stability and are expected to survive the dangers they will most likely face. However, further advancements in simulations have made it possible to further increase the fidelity and accuracy of simulated casualties. Multiple time domain and, to a lesser extent, Computational Fluid dynamics (CFD) solutions have been suggested as the next “evolutionary” step for damage stability. However, while those techniques are demonstrably more accurate, the computational power to utilize them for the task of probabilistic evaluation is not there yet. In this paper, the authors present a novel approach that aims to serve as a stopgap measure for introducing the time domain simulations in the existing framework. Specifically, the methodology presented serves the purpose of a fast decision support tool which is able to provide information regarding the ongoing casualty utilizing prior knowledge gained from simulations. This work was needed and developed for the purposes of the EU-funded project SafePASS.

Keywords

• damage stability
• case-based reasoning
• machine learning
• flooding
• decision support systems
• SafePASS