Meta-model assessing the probability of exceeding the allowed acceleration limits, with the use of Bayes networks

Tomasz Hinz, Maria Acanfora, Jakub Montewka, Przemysław Krata, Jerzy Matusiak

Research output: Contribution to conferencePaperScientificpeer-review

Abstract

Two stability related accidents are especially relevant to container ships, mainly due to their flared hull shape and cargo stowing patterns. They are parametric roll and synchronous roll, both resulting in excessive acceleration. Moreover, the actual loading conditions, which often vary from the declared one, and the expected weather conditions, remain uncertain and thus difficult to describe in a deterministic manner. Therefore, we present a concept of a causal probabilistic model suitable for the stability-related safety assessment of a container ship. The two modes of resonant rolling are studied, accounting for the pertinent elements of uncertainty. The model structure is constructed on the basis of data obtained by means of a series of simulations with the use of 6 DoF the state-of-art ship motion model called LaiDyn. For a selected container ship type C11 we adopted typical loading case with the KG fluctuation. Subsequently, the obtained data are organized into a probabilistic meta-model with the use of Bayesian learning techniques. The model sensitivity is examined concerning the weather conditions, which bear a significant amount of uncertainty in the day-to-day operations of container ships.
Original languageEnglish
Pages283
Number of pages290
Publication statusPublished - Sep 2018
EventInternational Conference on the Stability of Ships and Ocean Vehicles - Kobe, Japan
Duration: 16 Sep 201821 Sep 2018
Conference number: 13

Conference

ConferenceInternational Conference on the Stability of Ships and Ocean Vehicles
CountryJapan
CityKobe
Period16/09/201821/09/2018

Keywords

  • excessive acceleration
  • Bayesian Belief Networks

Fingerprint Dive into the research topics of 'Meta-model assessing the probability of exceeding the allowed acceleration limits, with the use of Bayes networks'. Together they form a unique fingerprint.

Cite this