This thesis examines the application of Evolutionary Algorithms for
ship hull form optimization. In addition the thesis briefly examines
the use of an Artificial Neural Network for predicting a particular
hull attribute.
A method is developed where different chromosomes are combined to
model the hull. A matrix chromosome is used to model the hull offsets.
A one-dimensional array or single chromosome is used to model the
principal parameters. The method allows both optimization of the
principal parameters to obtain an initial design satisfying the
requirements as well as concurrent optimization of the hull form
to minimize resistance and maximize other performance attributes
of seakeeping and stability.
Hull form optimization is conducted using hydrodynamic evaluations
of key performance attributes. In particular, performance in terms
of ship resistance, seakeeping and stability are evaluated. The design
methodology uses multiple objective optimizations and a novel
multiple objective optimization technique is developed.
The search for potential designs uses evolutionary algorithms to
optimize both the hull form and determine the principal parameters
satisfying the design requirements. A multi-species genetic algorithm
is developed to enable competition between alternate hull forms.
In order to obtain a reasonable approximation of the resistance, some
modifications of classical linearised thin-ship theory is utilized.
A particular problem for vessels with low length to beam ratio and
with transom sterns is investigated. In additional to resistance the
candidates are evaluated in terms of seakeeping performance. Seakeeping
is evaluated using a hydrodynamic evaluation of the hull forms in a
regular seaway. A two dimensional strip theory analysis of seakeeping
provides the input to develop a vertical motion seakeeping index.
With respect to stability an analysis of vessel candidates is conducted
using a regression based formulation and an artificial neural network.
A database of typical candidates is required to provide data on vessel
attributes that are required for training the neural network. The
determination of the center of gravity or KG is then used as input
for the seakeeping evaluation as well as to satisfy the constraint
for a maximum KG.
Key Words: Hull Form Optimization, Genetic Algorithms, Neural Networks,
Hydrodynamics, Ship Design.