Predicting Collision Damage and Resulting Consequences

Erik Sonne Ravn, Peter Friis-Hansen

    Research output: Contribution to journalConference articleResearchpeer-review


    This paper presents an Artificial Neutral Network (ANN)that is trained to predict the structural damage in the shipside resulting from ship-ship collisions. The input to the ANN is the absorbed energy, the length of the involved ships, the draught of the struck ship, and the angle of collision. The predicted output is the size of the hole (or holes in case of striking ship as a bulbous bow), which is given as the dimensions of a box. The ANN for damage prediction is used in connection with the risk evaluation of a selected navigational area, where the cost related to oil spills from tankers is estimated. It is proposed to use standard type risk profiles of lognormal typr for the cost of oil spill corresponding to a given operation time. The lognormal type risk profiles are particular useful since these only make use of mean and coefficient of variation of the single event loss and therefore are insensitive to variation of distributional assumptions of difficult tangible single event losses. The paper compares the accumulated loss over time resulting from oil spills in a given navigational area. The considered area is also evaluated in a future situation where the majority of all oil tankers have double hull. The expected socio-economic gain discussed. The ANN and the simulation procedure of oil spill cost are implemented into Excel, which can be downloaded at
    Original languageEnglish
    JournalSociety of Naval Architects of Japan
    Pages (from-to)7-16
    Publication statusPublished - 2004
    Event3rd International Conference on Collision and Grounding of Ships - Izu, Japan
    Duration: 25 Oct 200427 Oct 2004
    Conference number: 3


    Conference3rd International Conference on Collision and Grounding of Ships


    Dive into the research topics of 'Predicting Collision Damage and Resulting Consequences'. Together they form a unique fingerprint.

    Cite this