Ship design for damage survivability

Abstract

This thesis presents a new set of methods to assist the process of ship design for safety with particular reference to collision damage. The study has two principal objectives: " investigations into subdivision aspects of passenger ships to improve their overall survival index " investigations into the subdivision of oil tankers in order to improve the effectiveness against spillage in the event of collision damage. In order to investigate the ship subdivision aspect a damage stability model was needed. A pre-requisite for developing the damage stability software was a robust but flexible method to define the hull and the compartments of subdivision. B-splines have been a popular representationatl ool in computer aided design over the past three decades.T his method, though more complex than other spline techniques such as cubic splines, was adopted with a fourth order basis function in this work. A complete set of spline manipulation libraries and associated numerical solvers were developed for this purpose. In addition to this, a method to define the intersection between the hull and the waterplane in the form of a closed B-spline curve for any given orientation of the vessel in terms of -heel, trim and draught was developed to aid the damage stability calculations. Though the earlier regulations stipulate fixed trim assessments to ease the computational process, it is clearly unsatisfactory and research has confirmed this to be a flawed approach. Free trim calculations on the other hand require an iterative and time consuming process to arrive at the equilibrium trim position for each heel angle. Pawlowski proposed a new method for the stability calculations of a freely floating rig when the unit is arbitrarily orientated to the wind direction. It uses the Euler theorem on the properties of equivolume waterplanes to arrive non-iteratively at the new inclined position. This theory was adapted for use in damage stability calculations and was numerically tested and proved to be sound. Damage stability calculations, though combinatorially large, are also inherently parallel. Parallel Virtual Machines (PVM) is a Message Passing Interface (MPI) developed jointly by ORNL, University of Tennessee, Carnegie Mellon University and the Pittsburgh Supercomputing centre. PVM enables a "virtual configuration" so that a collection of serial, parallel and vector processing machines appear as one large distributed memory computer. PVM was compared with another MPI called Network Linda where the advantage of PVM's user controlled message passing was demonstratedP. VM was used to implement the MJMD Distributed Memory paradigm to exploit this inherent parallelism in damage stability calculations and to obtain speedups. A systematic exploration of the search space for this design problem involves the generation of a large number of internal subdivision configurations. This, coupled with the fact that the design space was multimodal in nature made it suitable to the application of a class of heuristic search algorithms called Genetic Algorithms (GA). A brief description of the mechanisms behind GA is presented along with their mathematical basis in the form of two theorems: the schema theorem and the building block hypothesis. Various techniques for solving constrained optimisation problems with GA was explored. The penalty function method was found to be the most suitable and was finally adopted. The above techniques were applied to the optimisation. of the internal subdivision of passenger ships and cargo ships, oil tankers in particular. For passenger ships, the nature of the 's'-factor formulation on the local index was shown. The multimodal nature of the subdivision problem was highlighted and a GA was used to investigate the optimal subdivision characteristics of the vessel. The 's' factor formulation for cargo ship rules is different to that described by the A. 265 set of regulations for passenger vessels. In addition, the cargo ship rules describe a factor V which accounts for the probabilities of vertical extents of damages. However this formulation does not assign any credit for horizontal subdivision below the waterline. Data on vertical extents and vertical location of damages for cargo ships was collected and analysed in earlier studies done at Newcastle University. This data was used to develop a probability function akin to that developed for the longitudinal extent and longitudinal location so as to give credit for any horizontal subdivisions. The principal objective of this part of the study was to explore the search space for subdivision configurations that would minimize net oil outflow.