Integrating Parametric Hull Generation into Early Stage Design

Marcus Bole, Graphics Research Corporation Ltd1

Byung-Suk Lee, University of Strathclyde-Glasgow

Introduction

The concept and initial design phases of a ship can be highly dynamic periods where main features
and dimensions are subject to frequent change in search of a good balance between often conflicting
requirements. The duration of this design period varies depending on ship type and owner requirements. Specialist one-off designs may require tailored engineering solutions. The hull form affects
hydrodynamics, production costs, and operation of the ship. Modern integrated ship design tools primarily focus on the analysis of hull performance rather than providing capabilities to investigate the
design space. While analysis is important, designers increasingly demand better hull surface definition
to allow ranges of solutions to be investigated easily before proceeding to detailed design.
Given the complex nature of the ship system, the design tools used to produce a solution must
address a wide range of engineering disciplines. In addition to analysing compliance with owner and
regulatory requirements, a successful tool must allow the user to investigate the design space adjacent
to a solution to identify optimisations or alternative approaches which may offer better performance.
As a platform supporting the entire ship system, the hull is the single most important component
in any vessel. Its form affects the hydrodynamic and structural performance of the vessel during
operation and the ease of construction when in production. Modern integrated ship design tools
primarily focus on the analysis of hull performance rather than providing capabilities to investigate
the design space. While analysis is certainly extremely important, the approaches used to assess hull
performance have matured and designers now need improvements in hull surface definition to allow
ranges of solutions to be investigated easily before committing to the detailed design phases.
2

Mathematics in Hull Form Representation and Design

While the mathematical representation of hull surfaces can be traced as far back as Chapman
(1760), designers preferred to use practical methods, battens and weights, to control the shape of
hull surfaces until the introduction of desktop computers and CAD software tools became affordable
and effective to use in a commercial environment. The use of mathematical representations of hull
surfaces was initially driven by the needs of analysis techniques, particularly hydrodynamics, rather
than designers. Taylor (1915) developed use a polynomial approach which assisted in the development
of a series of hull forms based on an initial parent that could be used to investigate the effect of different
form parameters on resistance. Benson (1940) extended this technique so that it could be quickly
used to generate waterline and section curves for use in the design office.
High-degree polynomials require many offset data points to produce acceptable shapes and introduce complexity in manipulation. For analysis such as seakeeping, an alternative hull representation
based on conformal mapping techniques was developed. These approximate the hull sections by simpler
geometric forms permitting quasi-analytic hydrodynamic solutions. Von Kerczek (1969) developed one
of the first computer based hull design systems based on this technique where the user could use a
light pen to manipulate point offsets which would change the shape of the hull surface.
With a mathematical relationship between hull shape and performance, it was not long before
attempts were being made to develop techniques which would produce the hull form geometry based
on parameters governing geometric and performance characteristics. Kuiper (1970) introduced a
technique loosely based on Bensons approach which rather than base the hull representation on
offset data used initial parameters describing hull characteristics used to generate longitudinal form
curves (waterline, profile, section area curve) or influence functions used to generate individual section
1

Froude Rd - Haslar Marine Techn. Park, UK Gosport P012 2AG, mb@grc-ltd.co.uk

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

115

geometry. This approach was rather successful and used by many subsequent techniques. However,
there were still limitations in the mathematical functions used to represent the hull surface itself in that
they were just not flexible enough to represent the range of hull surface shapes. Reed and Nowacki
(1974) combined the use of the conformal mapping technique below the waterline and polynomial
functions above to make a small improvement and allow flare, but limitations still remained.
While these developments were taking place, computer systems became affordable for large commercial organisations and research centres. Investigation into using these systems for both design and
manufacturing began with particular interest from the automotive industry. New freeform parametric
curves like Bezier curves, Bezier (1972), and B-Splines, Riesenfeld (1972), were significantly more
flexible and easier to work with than either conformal mapping or pure polynomial techniques. Hull
surface design techniques were adapted to work with these new representations, Creutz and Schubert
(1978), Munchmeyer et al. (1979) greatly improving the results. However, with the introduction of
parametric surface counterparts of the curves, particularly NURBS, many users found that the representations alone could be used for hull surface design without any supporting technology. It was
good enough to manipulate the surface definition point until an acceptable surface is obtained in an
iterative process analogous to using batten and weights.
3

Computer Aided Hull Design in Ship Design Tools

When cheap personal computing became available in the 1980s, the effectiveness of hull design
software took a great step forward when combined with parametric surface representation techniques
such as NURBS. Prior to this, computer aided hull design tools had been no more than functional
electronic versions of the paper lines plan. The introduction of windowed graphical user interfaces and
interactive peripherals, such as the mouse, allowed developers to produce CAD tools that could operate
consistently across a wide range of low cost computing hardware. As a result, NURBS surfaces have
become the most popular technique for describing the shape of hull. The surface definition (control
vertices) can be interactive manipulated using the mouse and programmers with even novice experience
can easily develop the algorithms required to produce the representation.
Before intuitive interactive design became one of the most important advantages of parametric
surface representations, the initial benefit from the introduction of techniques such as NURBS was
the ease and accuracy that could be achieved in production activities. The mathematical descriptions
of component parts could be used to directly control tools such as cutting machines from a centralised
design database rather than have this control information generated separately by reviewing the many
drawings produced for the ship. Production is the most labour-intensive activity in shipbuilding.
Therefore ship design tools have concentrated on providing the best solution for the detailed stages
of design, seeing there the greatest potential savings in initial cost. However, although production is
the most labour intensive activity, it is during early design phases where the designer has the most
influence on the through life costs of a vessel.
Today, while many shipyards still rely on 2D methods, the main focus of ship design tool development is in the use of database or PDM (Product Data Model) technologies as a centralised data
repository for all design definition and associated analysis results. On projects which require many
concurrently working engineers, the design needs to be kept updated by allowing changes to automatically propagate across all affected components and for any infringements to be flagged up. For
example, a change in the location of a bulkhead must lead to an update to the definition of the adjacent
compartments which will subsequently lead to changes in the damage stability, etc. To achieve this,
the design database must retain the topological relationships between different components as well as
geometrical definition. While almost all geometric components in a model can be updated because
their individual influence on the design is small and definition simple, the shape of the hull cannot be
updated so easily as existing design tools do not possess the ability to topologically describe how hull
form shape changes, preventing even the simplest of hull surface definitions from being updated.

116

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

By focusing on the management of the complete design model, the development of these integrated
tools has focused on the detailed phases of design to the detriment of early stage design capabilities
where a coarse definition of the ship is preferable. However, as most of the analysis techniques are
just as applicable at the early design stage these tools cannot be disregarded, being one of the reasons
why many of these tools are used throughout the whole process. Therefore, in order to improve the
hull design process, the best approach is to adapt existing hull surface definition techniques allowing
them to respond to changes commonly found in integrated design tools. Moreover, this needs to be
achieved without dictating only a detailed level of design. In fact, the hull definition technique needs
to be able to respond to a greater range of changes at global and medium levels as well as at detailed
scale. As a result, integrated design tools should become more amenable to early stage design allowing
a single platform to support the whole design process.
4

Current Hull Design Techniques

There are several different approaches used to interact with the hull surface depending on whether
an acceptable definition exists or on what type of design process is to be employed.
4.1

Manual Development of New Hull Surface Definitions

To develop a new hull form, the data defining the surface representation must be built up and this
is normally achieved using an iterative process where the definition is manually modified until the
surface is fair and it fits the design requirements. Depending on the scale of the design there are two
approaches that can be used:
Direct Manipulation of Surface Definition Data
For small vessels with relatively simple surface, it is possible to manipulate the raw surface
definition (control polygon) directly. This approach is most favoured in popular low cost design
tools as the algorithms used to generate the hull surface are simple to implement and control of
the definition data is intuitive. However, as the scale of the design increases, the time required
to develop the hull surface increases more significantly and the process becomes very tedious.
Furthermore, without expert knowledge, specific features cannot be accurately introduced into
the surface. Despite being the easiest and most artistic technique, its flexibility makes it the
least productive. Examples of commercially available tools which allow the user to manipulate
the surface directly are Autoship and Maxsurf.
Structured Hull Definition Techniques
Rather than control the mathematical surface definition, tools used to develop more complex
hull forms often generate the surface definition from a set of curves. The curves are used to
create a net which the design tool interpolates to generate the hull surface. In addition to
definition locations which the surface will interpolate, curves can include attribute information
used to constrain the surface, by applying specific tangent along an edge or by creating a corner
(knuckle) along the curve for example. Detailed features can be easily constructed within the
hull without affecting other areas of the surface by using irregular curve networks. Although this
approach is much more productive, development of good curve networks requires experience and
it can be difficult to change hull shape if the structure of the curve network must be changed.
This is not necessarily the fault of the technique as improvements in the user interface can greatly
enhance the designers experience. Examples of commercially available tools with structure hull
surface definition are NAPA, PIAS, EzHull, Lee et al. (2001), and to some extent Multisurf,
Letcher et al. (1995).
4.2

Transforming Existing Hull Surfaces for New Designs

As the process of developing a new hull form definition can be lengthy and tedious, designers are
often reluctant to modify the surface, particularly if the design has been proven or has had extensive
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

117

analysis. In this situation, hull transformations allow the surface to be consistently altered minimising
the need for further fairing. The origins of these transformation can be traced back well into the Lines
Plan era where time spent on addition fairing could be costly. Basic affine transformation can be used
to change the overall dimensions of the hull surface while functions more specific to hull design can
be used to change the volume distribution, Lackenby (1950). Due to their relative simplicity, these
functions are unable to respect the shape of many surface features and can introduce undesirable
distortions if transformed significantly from the original design. Despite this, these functions are
considered core tools for any practical ship designer as they offer a relative cheap way of reusing
existing hull forms and are generally available in most hull surface design tools.
4.3

Automatic Generation of Hull Surface Definition from Form Parameters

Neither direct surface manipulation nor hull form transformation are particularly efficient tools.
The designer spends a considerable time to continually check the quality of the surface and remove
undesirable features. Hull surface generation offers an alternative by aiming to construct a fair surface
from a predefined set of characteristics. Parametric hull generation techniques were popular developments until the introduction of parametric freeform surfaces like NURBS which allowed the same
hull representation from concept to production design. More recently, increased pressure to faster
turnaround design proposals and the potential for automatic optimisation of the hull surfaces has
rekindled interest in these techniques. Parametric freeform surface representations have matured.
NURBS have become the industry standard of modelling CAD surfaces. These representations can
now be wholly embraced and incorporated into modern parametric hull generation tools. The Friendship modeller, Harries (1998), is an excellent example of a tool based on form curves but ultimately
producing a NURBS hull surface. In combination with hydrodynamic analysis and optimisation tools;
it produces hull surfaces with particular performance characteristics.
There are other recent examples of parametric hull generation techniques not using freeform parametric surfaces. Jorde (1997) developed a hull generation technique based on Microsoft Excel demonstrating that anyone who wants to develop their own technique has access to tools which can support
the functions required. Lowe (2005) updated work by Bloor (1990) combining resistance optimisation
through the use of genetic algorithms based on a hull form defined by partial differential equations
(PDE). Of the many different parametric hull generation techniques, few are used in actual ship design.
In order to have a greater understanding of the limitations, two different hull generation techniques
were developed as precursors to the work here:
YachtLINES is a single cubic B-Spline surface yacht hull generator based on 19 geometric parameters. It follows the basic approach taken by many previous techniques using longitudinal form
curves from which section shape is generated. Form curves are defined using B-Spline curves and
an iterative approach is employed to modify control vertices until the desired hull properties are
reached. A final NURBS hull surface representation is generated by performing a longitudinal
fit to the control polygons of each section. Recent updates to this technique to constrain input
parameters to ranges capable of being produced by the technique and the addition of bracketing
of the iteration procedures have resulted in very robust performance, Fig.1.
ShipLINES, Fig.2, produces a B-Spline surface of a single-screw cargo ship hull form with and
without bulb based on 25 geometric parameters. Most of these are used to define local appendages such as the bulb and shaped skeg. The control polygon of the hull surface is specified
directly around the areas of the stem, midship section, transom and skeg. The rows of the
surface are blended in between taking account of the parallel middle body. Unfortunately, this
construction technique and the requirement to produce both bulb and skeg in the surface place
a significant constraint on the surface definition such that it is no longer possible to control the
hydrostatic properties independently of the other input parameters. Despite this, the surfaces
produced by this technique are reasonably good as the effect of the high level of constraint results
only in a reduction in the range of hull shapes that can be produced.

118

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

a) Initial parametric hull

b) Clio
c) 1m RC Yacht
by Philip Craven
by David Larsen
Fig.1: Three examples of hulls generated YachtLINES

a) Everise Glory
b) Thanh Da
Fig.2: Two hull forms generated by ShipLINES; courtesy of SMIT Singapore
The primary aim of hull surface generation techniques is the production of a fair form from a set of
numerical parameters providing the user with an almost instantaneous way of creating a new design
with the correct characteristics. However, as both these hull generation prototypes demonstrate,
hull forms can have many different characteristics and it is very difficult to find a mathematical
formulation which will accurately produce appropriate surfaces with even a small range of variation.
The dilemma faced in developing these tools is either to produce a wide range of generic hull forms
which must be modified manually once generated or produce detailed surface definitions with similar
characteristics and limited variation. Specifying a hull form using numerical parameters is not always
easy. Parameters such as dimensions and volumetric coefficients are obvious inclusions but additional
parameters are required to complete the definition and allow some variation in the range of hull forms
that can be generated. Often, numerical parameters must be used to define characteristics which
may be difficult to describe and may control features where it is more intuitive to interact directly
with the surface definition. Furthermore, detailed features such as bulbs or shaped skegs may require
many parameters dominating the definition making the technique difficult to use. An appropriate
solution would be to integrate hull surface generation technique into a tool that also allows interactive
manipulation. However, as it is very difficult to design a mathematical surface generation function
which can also adapt to interactive user customisation of the hull surface, the change from a generated
surface to one which can be manually modified is both unacceptable and irreversible. Consequently,
designers have found hull generation technique difficult to integrate into their approach and opt to
use the other hull design techniques.
All three hull surface design techniques have scope for improvement. But rather than address
the disadvantages though separate technical improvements and retain the existing techniques, an
alternative approach will be taken to identify the capabilities required by the designer allowing the
investigation of innovative engineering solutions with minimal overhead resulting from the operation
of the design tool.

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

119

Approaching the Problem from the Designers Perspective

Research to improve hull surface design techniques is still active, as the methodology behind many
accepted tools has not changed for some time. Unfortunately, most new developments do hardly
consider the needs or opinion of the users. We take here an alternative approach to formulate a
framework for future hull surface design platforms by considering:
the process used by designers to synthesize new hull forms
the capability of existing surface representation techniques
the limitations enforced by current working practice (data exchange, capabilities of other CAD
tools, etc.)
A designer will have a good mental picture of the hull surface when developing a new design to
the extent that, even without the assistance of a CAD system, a representative sketch, Fig.3, can
be readily produced. Of course, the sketch will not have any accuracy but it may illustrate key
characteristics of the design such as knuckle lines, flat areas, bulbs and propulsion arrangements etc.
Consequently, the designer associates most closely with the features of te hull surface more than
any other characteristic of the hull form, Fig.4. Looking more closely at the contents of the sketch,
the arrangement of hull features forms a topology of surface shapes and could assist in the design
process making it more manageable by effectively dividing the surface into separate parts which have
some independence from each other. Furthermore, there are only a limited variety of different types of
hull form and the shapes within the topology can be easily described. This offers the possibility that
for a given topology, the design tool could develop much of the shape of the surface representation
automatically.

Fig.3: Basic hand sketch of a ship hull form

Fig.4: Hand sketch with rendering and text

descriptions of surface shape

Numerical control of principle hull form characteristics has always been an attractive concept and
has resulted in continued interest into parametric hull generation. However, as these methods require all characteristics to be defined numerically, description and control of shape has always been
unsatisfactory. Ultimately, the designer wants to be able to control dimensions numerically but still
be able to control shape through interactive manual manipulation. With an underlying topological
representation of the hull form, there is an opportunity to use numerical control to change dimensions
of the hull form with the reassurance that the topological definition will maintain consistency between
the features of the surface.
One of the underlying limitations of current hull definition techniques is control of detail. Direct
definition of the surface representation requires high resolution information about local details while
parametric hull generation techniques often ignore important local features completely. Consequently,
120

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

there are often problems when moving through the design process requiring the hull surface to be
redefined. Definition of the hull form topology could potentially assist this process as it records the
underlying structure of the surface. The surface within a particular zone could be refined using the
previous definition without affecting other areas of the surface. Furthermore, if some areas had been
originally defined using generation techniques then as the definition progressed through subsequently
more detailed design phase these areas could be converted to geometric definition to provide the
designer with more local flexibility. As a result, the technique is capable of adapting its resolution to
the required level of detail for a particular point in the design process.
An understanding of the common hull form surface shapes and control of geometric definition and
numerical parameters appears to be primary requirements which would enable a hull design tool to
be used in an early stage ship design environment, while at the same time having the flexibility to
produce a range of different designs by providing the designer with artistic control of shape which
does not require any significant fairing. Having formulated the concept of a hull design tool which
solves some of the limitations of existing techniques and aims to fit the needs of the design process
more closely, the next stage is to look at the technical aspects behind the implementation of a tool
which follows this philosophy.
6

Technical Considerations for a New Hull Surface Design Tool

Just as it is important to review the hull design process when developing a new technique, it is
equally important to consider the practical aspects of the technology used within any new methods
and the environment in which to the tool will be used:
Mathematical surface representation: There are many different mathematical surface techniques
used throughout the CAD industry although NURBS is by far the best supported both as an
intuitive surface design tool and for transferring data between different systems.
Access to the surface flexibility: Experience from parametric hull generation demonstrates that
closed approaches, where there is limited practical opportunity to change the way the surface
is constructed, greatly constrain the designer. A preferable alternative is to employ an open
technique by providing tools which can be optionally used to constrain the surface. These
constraints may be initially applied by the tool to attain the correct surface shape according to
the defined topology but may be later removed by the designer when access to the full extent of
the surface representations flexibility is required.
User familiarity: Traditional techniques to control surface shape allow users of existing tools
to use the new approach with a minimum of training. The user interface allows the designer
to control the hull surface ab initio by providing intuitive tools for flexibly controlling and
accurately positioning geometry. Ideally, curves should be used to control the hull form as they
provide clearer feedback of shape. Surface control polygons, while familiar to most designers,
often become distorted and can require a good level of interpretation to understand how an
individual control point affects surface shape. Numerical parameters should be clearly labelled
so the designer understands exactly what it controls.
Procedure and sequence: In addition to good technical capability, the most effective tools emphasize a preferred procedure that should be followed when assembling a surface definition.
While subtle, a procedure makes the process of creating a surface definition more deterministic
as the designer always knows what steps are still to be taken and the implications of modifying
a particular aspect of the surface.
Integration with other tools: While there are many tools which provide integrated solutions
it is common for many different tools to be used during the design process. Therefore, any hull
surface design tool should allow data to be exchanged between other systems using commonly
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

121

supported representations and with a high degree of accuracy. In practical terms, this
means supporting bi-directional exchange of data through file formats such as IGES and DXF,
with consideration for future developments of STEP and XML based formats when they mature.

Realisation: Combining Functional Aspirations with the Technical Reality

In existing hull design solutions, the software has a static role in the design process providing an
instance of the surface representation. The user controls shape by manipulating the representation
definition structure taking on the sole responsibility to ensure that the surface is correct. However,
a tool which assists the designer in the surface development process as well as supporting all the
technical requirements of a hull surface design tool requires a more innovative solution. Certain
functions will require the tool to analyse the surface representation to assess the best way to modify
the definition to the users demands. By introducing a framework that interfaces between the user
and the resulting surface definition, the software can control what aspects of the surface can be
manipulated and introduce specific features needed in the hull design process. Curves offer the best
means of controlling hull form shape for the following reasons:
Can be used to define the topology of the surface, marking the boundaries between the different
features of the hull form
More intuitive control and representation of shapes within the hull form
Functions used to modify the shape of the hull form will be the least complex when implemented
for curves
Tools that constrain curve shape are understood by users more easily than if implemented directly
on the surface definition
7.1

Simplifying surface definition and reducing modification

In order to speed up hull form definiiton and to allow the surface to be modified consistently methods
of reducing the amount of definition and controlling surface shape with respect to the topology need
to be introduced into the framework. Rather than relying on the user to correctly position all curves,
a collection of tools which impose geometrical constraints on curve definition can be used to produce
shapes commonly found in hull forms. These constraints will maintain shape relationships between
control points and between curves forming a hierarchy based on the underlying topology of the surface.
Consequently, changes to individual curves will need to propagate to others maintaining the correct
hull form topology.
7.2

Addressing Parametric Control of the Hull Surface

Using the topology of the hull form and the hierarchy of constraints, introducing parametric control
to a geometrically defined surface is now easier. In traditional parametric hull generation techniques,
the topology of the hull form is encoded in the complex mathematical formulae used to produce the
surface making it very difficult to change how the hull is shaped. Having topologically represented hull
form shapes, these mathematical procedures no longer have to consider the implications of changing
shape, as this is maintained by the applied constraints, and they have only to identify which aspects of
the hull form to transform, which can be achieved by interrogating the topology. Parametric control
can now be introduced to the framework as an additional set of tools. Individual parameters can be
measured from locations in the surface definition identified by analysing the topology. Subsequently,
a transformation function can be invoked when a parameter is changed by a user, which modifies
specific characteristics of the hull definition.

122

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

7.3

Introducing Independently Defined Local Features to the Hull Surface

The introduction of local features to the hull surface, such as bulbs or propeller skegs, often requires
increased definition in the area of interest to allow the feature to be modelled. However, this may only
be achieved by increasing the level of detail in definition across greater parts surface making further
modifications to the hull form difficult, as much more data must be kept fair. As the framework is
effectively managing the surface definition, an alternative approach can be introduced which allows
features of the hull surface to be added sequentially by refining only the generated surface definition.
By handling the introduction of hull features separately, the definition of the initial hull surface and
subsequent features remains independent so the surface is always just as easy to modify and the
designer can add, remove or change the shape of any features of the hull without having to make any
irreversible changes to a unique surface definition.
8

Implementation of a Concept Demonstrator - IntelliHull

IntelliHull, Bole (2002), was developed as a simple demonstrator of some of the proposed functions
using a single NURBS surface and focusing on ship hull forms with and without bulbs. IntelliHull
implements each aspect of the design tool discussed in the previous section as follows:
8.1

Framework

One of the difficulties faced when representing a ship hull form using a single NURBS surface is
the large number of vertices in the control polygon which each have three independent parameters
(coordinate components) that must be specified to produce the surface. While some techniques have
used goal seeking techniques which manipulate each individual vertex, Sanderski (1998), a more
effective approach is to construct a geometric framework based on selected hull characteristics which
can be used to determine the location of the surface control polygon vertices using a predefined
set of rules. The ShipLINES prototype illustrated that a fair (i.e. smooth from the mathematical
perspective) hull could be generated by blending the vertices of the control polygon between the bow,
parallel middle body and transom if local features and appendages were ignored. This perhaps invokes
the idea that a fair control polygon will produce a fair hull.
The framework for IntelliHull is based on the idea of three separate areas or panels of the hull
surface representing the entrance, parallel middle body and run of the hull. In many ways this format
is similar to the arrangement used in both Tribon FORM and FORAN FORMG although neither
of these two techniques was originally designed to produce a NURBS surface. Each row of the hull
surface control polygon is associated with a B-Spline curve used to blend the shape of the hull, Fig.6.
In order to produce a fair hull surface, the curves must behave as a family with similar arrangements
of control points and discontinuities. A series of transverse definition curves are used to control the
blending curves. The vertices on each transverse curve define the location of an equivalent vertex on
each blending curve producing a coarser equivalent of the control polygon.
By arranging the transverse curves to form discontinuities in the shape of the longitudinal blending
curves with respect to the key characteristics of B-Splines, Rogers (2000), the three panels of the
hull surface can be exposed allowing full control of the panel boundaries, tangents and fullness in the
centre of each panel. To form a hull surface with entrance, parallel middle body and run, the definition
of many transverse curves are geometrically related to other curves in the definition. For example,
within the parallel middle body panel and associated tangent curves, equivalent vertices on each curve
must lie on the same line in the X coordinate direction to form the prismatic shape of the hull. By
constructing the framework entirely based on B-Spline curves, rules based on the key characteristic
can be defined and used to constrain and relate separate definition curves together to form a ship
hull surface. Consequently, in addition to manipulating the vertices of each definition curve, the tool
can provide a set of tools used to constrain definition curves producing the desired hull form shape.
It is possible to take this concept further. For example, if the characteristics of the hull surface can
be defined within the tool with some topological description, it can generate much of the hull surface
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

123

definition itself leaving the user to manipulate definition which controls the characteristic shape of the
hull surface.
To form a ship surface, IntelliHull needs between eleven and fifteen curves depending on what
characteristic features are desired in the hull surface. A Ro-Ro type hull, for example, requires
thirteen curves, Fig.5. Seven of these can be automatically generated from the definition of other
curves and further user assigned constraint tools can be used to simplify the definition of other curves
by assigning specific shapes to the definition creating, for example, the bilge radius.

Fig.5: Curves representing Ro-Ro hull

Fig.6: Blending functions at entrance region

Control Curves and Constraints

8.2

Control curves were highlighted as one of the methods designers associate best with for controlling
hull surface shape. The framework exposes definition curves but to generate an accurate ship hull
form the curves need to be placed in certain geometric relationships with respect to each other to
produce the discontinuities and tangents forming the panels. Such a system is not intuitive to use and
requires the designer to understand the theories behind B-Spline representations. Rather that rely on
the designer to place the definition accurately, the tool can provide features which assign these rules
to the hull definition in the form of constraints. IntelliHull provides a range of different constraints
which assist in the development of transverse definition curves.
8.2.1

Vertex Level Constraints

These are the simplest of constraints and provide the designer with the ability to create certain
shapes within an individual curve by assigning rules to a contiguous set of control points. The rules
constrain the internal vertices of the selection using the known properties of B-Spline curves. The
variations available are:
Straight: Internal control points are constrained to lie linearly and at uniform intervals between
the two end points, Fig.7. This constraint is used for developing flat of side and flat of bottom
in the midship section or for constructing straight sections in the stem or transom. It can be
used generically in all control curves.

Fig.7: Straight segments

Blend: Internal control points are constrained to lie at uniform intervals on a cubic spline formed
from tangent information determined at the end points of the selected range of control points,
Fig.8. The tangents are determined from neighbouring constraints or from adjacent vertices if
there is no other information. The blend is tuned so that for direction changes of 90 will result
as close as possible to a circular arc. The blend can be used to form the bilge radius in the
midship section or to form curved shapes in the stem and transom.
124

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

Control points fitted

cubic spline

Fig.8: Blended curve segments

Knuckle: According to the known properties of B-Spline curves, a cusp can be formed in a cubic
B-Spline by collocating three adjacent points. Knuckle constrains two vertices adjacent to lie at
the location of the selected vertex, Fig.9.

Fig.9: Knuckle constraint

8.2.2

Curve Level Constraints

Curve level constraints assign rules which apply to all the vertices on a single curve, and subsequently all of the blending curves at that location along the hull. The two examples implemented in
IntelliHull are:
Plane: The plane constraint forces all the vertices on a curve to lie on a plane, Fig.10. This
is useful for simplifying the definition, and hence manipulation, of planar curves such as the
midship section and stem. It can also be useful for defining inclined transoms. The benefit
of assigning this constraint is that the curve can be manipulated by changing the definition of
the constraint rule by modifying, for example, the location of the plane (to move the midship
section) or by changing inclination (to vary a transom).

Reference
Curve

Constrained
Curve

Fig.10: Plane constraint

Fig.11: Offset constraint

Offset: Offset constrains certain coordinates of each vertex on the curve to be the same as the
corresponding vertex on a referenced curve. Consequently, it appears as though each vertex is
fixed to a line through its equivalent on the referenced curve although the user is free to modify
how far along the line the vertex is located, Fig.11. The constraint supports relationships
functioning in each of the three primary coordinate directions or can extract the vector between
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

125

the initial points of both curves. All curves which form the parallel middle body can use this
constraint so that they acquire the same transverse shape as the midship section. Their shape
is updated as the midship section shape is modified.
8.2.3

Automatically Generated Curves

Certain definition curves affect the shape of the surface strongly. Examples are curves that form
tangents and control blending in the middle of panels affecting prismatic coefficient. In most cases
it is easier to control these curves using geometric rules and numerical parameters rather than user
manipulated control points. Consequently, these definition curves can be automatically assigned to the
hull definition by the design tool, but later manually customised by the user. IntelliHull implements
the following three curves which are generated parametrically.
Tangent Control Curve
The direction and magnitude of tangents in B-Splines is determined by the arrangement between
the outer two control vertices of the control polygon or vertices on either side of a discontinuity.
In the framework, tangents in the hull surface at the bow and transom are controlled by the
relationship between the curve at the panel boundary and adjacent curve. The vertex and curve
constraints can be used to define this relationship but the longitudinal blending curves need to
reflect this information and allow the surface to blend into the tangent. This is achieved by
assigning the same tangent direction and magnitude to the blending curves but using the user
tangent definition curve rather than the boundary curve as the origin of the tangent, Fig.12.
Bow
Curve

Bow Tangent
Curve
Tangent Control
Curve

Blending Curve Control Polygon

Blending Curve
Tangent Control Curve (Generated)

Hull Surface Control

Polygon (Row)

Bow Tangent Curve (User Defined)

Bow Curve (User Defined)

Fig.12: The tangent control curve adds tangency into each blending curve based on the user
curves defining the end of a panel in the surface.
Left: Control points from each transverse definition used defining end of blending curve.
Right: The arrangement is repeated for each blending curve
Flat Control Curves
At each end of the parallel middle body, the direction of the blending curve tangents are known
(they lie in line with the parallel middle body) as indicated by the topology of the hull form.
Thus it should not be necessary for the user to explicitly add this information. The approach
used for the tangent control curve is extended by generating both the curves which control the
tangency in the hull surface and the tangency in the blending curves. The vertices of both
curves lie on lines extended through the vertices of the midship section. The only thing left
to be determined is the magnitude of the tangent. While experimenting with the definition,
the most aesthetically pleasing effect was achieved if the distance between the boundary curve
and tangent control curves was equivalent to a uniform distance normal to the panel boundary,
Fig.13. The construct has to be translated into control vertex positions on lines formed through
the corresponding vertices on the midship section definition using an equivalent operation to
offsetting in CAD tools. The offset is exposed to the user as a controllable numerical parameter
126

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

as a percentage of the length of the entrance or run panels depending on which end of the parallel
middle body is being controlled.

Fig.13: Controlling magnitude of tangent at ends of parallel middle body

Volume Control Curves
The framework allows for control of panel shape between boundaries. The user can add further additional control curves. If none is provided the system will introduce an automatically
generated curve which can be used to control the volume and centroid of the underwater hull
surface. The goal of the volume control curve is to expose a single parameter which can be used
to control the shape of the panel so that it can exhibit the maximum and minimum volumetric
fullness given the shape of the hull at the boundaries. Volume control curves have a geometric
framework, Fig.14, of their own to allow the control of shape based on the following specification:
A control parameter for each vertex on the curve is introduced and based on the interval
[0,1] representing the minimum to maximum range of fullness.
The longitudinal location of each vertex is halfway between the corresponding vertices on
the adjacent tangent control curves at each end of the panel.
At maximum fullness, the transverse location of the vertex (y, z) is the same as the corresponding vertex on the tangent control curve at the boundary to the parallel middle
body.
At minimum fullness the transverse location of the vertex is the same as the corresponding
vertex on the tangent control curve at the applicable bow or stern boundary.
The location of the vertex varies linearly with the control parameter between these two
positions.
Volume Control
Curve

Volume control
curve vertex
Vertices on tangent
curves on Parallel
Middle Body

Vertices on tangent
curves at Panel
End (Bow)

Fig.14: Volume control curve

A schedule mechanism determines the control parameter of each vertex which varies the fullness
of the hull at the keel while keeping the shape at the deck constant, Bole (2002). Additional
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

127

complexities are introduced when considering the different shapes associated with flat of side
and flat of bottom.
Constraints are applied so that parts of the definition will conform to certain rules leaving other
parts for the designer to modify. All of the constraints are applied as active rules so that if any
part of the definition is changed the constraints are applied and the surface regenerated. This
requires the generation process to work in an ordered way. Constraints associated with a particular curve are managed and owned by that curve. The curve executes each type of constraint
based on a predetermined order. Once all user defined curves have been updated the generated
curves can be recalculated again based on a predetermined order. If one generated curve references another, the referenced curve will update first. Obviously, no circular dependencies are
allowed and the user is not provided with any options that would allow this to happen.
8.3

Parametric Control and Transformation of the Definition

One of the drawbacks of directly manipulating the mathematical surface definition is the number of
times individual vertices must be moved to achieve a fair surface after a change. Constraints minimise
this activity because the assigned rules can automatically update many vertices in one operation. For
specific changes to the hull surface, parametric modification can be used for changing measurable
characteristics of the hull surface. This can be implemented by using focused transformations which
modify specific parts of the hull definition and propagate throughout by the assigned constraints.
Values of adjustable parameters are measured directly from the hull definition and changes made by
invoking specific transformation routines. IntelliHull could be described as being an indirect approach
to parametric hull generation as it leaves the user free to manipulate geometry and create shapes
which may be difficult to describe numerically.
LOA
Parallel Deck
Parallel Deck Aft (PDA)

Parallel Deck Fwd (PDF)

Displacement (DISP, CP, CB, LCB)

Parallel Middle Body Aft (PMBA)

Parallel Middle Body Fwd (PMBF)

Parallel Middle Body (PMB)
LBP

Breadth (BWL)

Depth (D)

Draught (T)

Fig.15: Primary parameters used to transform the hull definition

8.3.1

Geometric Transformations

All geometric transformations can be described as either translation or scale operations, or combinations of the two. Both translation and scale operations can be applied to the entire definition of
a curve or to a selected range of vertices where one of the vertices will form an origin in the case of
the scale operation. Operations involving combinations of both scaling and translation can be used
to minimise distortion which may occur if a single transformation was used allowing sections, for
example, to retain the same bilge radius.
Midship Section Transformation
Parametrically changing the depth or breadth of the hull form invokes transformations which
primarily focus on the midship section but are applied to the other user defined curves. The
128

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

two transformations are designed to minimise the distortion that would occur if a single scale
operation was applied by combining translate and scale operations together, Fig.16. Before the
transformation is performed, the section geometry is checked to identify the vertices that should
be moved and those that should be scaled. For changing breadth, the vertex at the bottom of
the bilge radius is identified. Vertices inboard of this location are scaled while vertices outboard
are translated ensuring that the shape of the bilge radius is retained. For changing depth, the
vertex at the top of the bilge radius is identified and vertices above this location are scaled. Both
these transformations assume that no constraints are applied to the midship section and try to
retain the original shape. Any applied constraints update after the transformation to bring the
section definition in line with the rules assigned by the user.
Scale control vertices
flat of side about (a)

Scale control vertices

inside and translate those
outside of (b)

(a)

(b)

Fig.16: Midship section changed with a combination of scaling and translation

Parallel Middle Body and Deck Transformations
Transformations on the extent of the parallel middle body and deck work in a similar fashion to
those used to change the midship section except that only scale operations are used. IntelliHull
extracts the length of the parallel middle body and deck from the definition, but also provides
additional parameters which describe the longitudinal locations of where these measurements
were taken, i.e. the curves defining the forward and aft boundary of the parallel middle body
panel.
Measurement
Extent
Forward Location
Aft Location

Symbols
Parallel Middle Body Parallel Deck
PMB
PD
PMBF
PDF
PMBA
PDA

The curves are only transformed when the specific parametric measurement relating to the
particular curve are changed. For example, the curve defining the forward end of the parallel
middle body can only respond to changes in the values of the PMBF and PDF parameters.
When the user makes a change to either the extent of parallel middle body or deck, the change
is distributed between the location parameters at the forward and aft end by the ratio of the
location of midship section curve within the parallel middle body. The transformation of the
actual curves takes the form of a scale operation about selected vertices, Fig.17. To change
the end of the parallel deck, the vertex on the deck edge is shifted to the desired location and
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

129

then the vertices between it and the vertex closest to the midship section are shifted by scaling
longitudinal location by the ratio of vertical vertex locations in the flat of side. A change to the
parallel middle body is achieved similarly. The vertex closest to the midship section is shifted
by the desired amount and vertices either side are shifted by scaling longitudinal position by the
ratio of the vertex location in the flat of side or flat of bottom whichever is appropriate. Again,
if any constraints have been applied to these curves, vertex locations will be updated after the
transformation has been applied.

Fig.17: Transformation of PDF (left) and PMBF (right) by selective longitudinal scaling
Length Modifications
Changes in length are implemented as translations or scale transformations applied to the individual curves. Before any geometry is changed, the hull definition is first checked to identify if
the hull has an extent parallel middle body. Depending on whether the hull is being increased or
reduced in length the transformation will change the length of the parallel middle body before
performing any scaling transformations to avoid introducing distortions. Consequently, if the
hull is being reduced in length but there is not enough parallel middle body to absorb the change,
the transformation process will first translate all curves, from the midship section to the stem, aft
until the extent of parallel middle body is zero. Subsequent further reduction will be introduced
by scaling the definition about the origin (Aft Perpendicular). While this technique could be
described as being arbitrary, it does not matter to the hull generation process how geometry is
changed as the technique is not dependant on the implementation of individual transformations
and could work with any designed to make an appropriate parametric change.

Fig.18: Length changed by changing just parallel middle body (left) or whole definition (right)
8.3.2

Hydrostatic Transformations

One of the primary objectives of almost all parametric hull generation techniques is to produce a
hull surface with particular hydrostatic characteristics. ShipLINES demonstrated that if too much
attention was focused on the geometric aspects of hull shape, the ability to control hydrostatic characteristics is lost. The hydrostatic transformations used in IntelliHull work in conjunction with the
automatically generated volume control curves defined in 8.2.3 and allow the volumetric characteristics to be changed independently of any curves the user may have defined. Most parametric hull
generation techniques produce the hull surface by using a model of the hydrostatic characteristics.
However, the hull surface already exists in IntelliHull when hydrostatics transformations are invoked
and mean that the response of the definition to any changes must be used to guide the transformation
process.
The transformations need to be applied iteratively because the only accurate model of the hull
characteristics is the actual surface definition and far to complex to make any attempt at solving
directly. YachtLINES uses iterations to manipulate B-Spline curve definition and demonstrates the
130

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

feasibility of the approach. However, the iterative techniques used were nave and the results were
slow and far from robust. (YachtLINES has since been updated to incorporate the improved iterative
technique used in IntelliHull and is now much faster and significantly more robust.). IntelliHull uses
several approaches to improve performance. Firstly, the parameters that drive the shape of the volume
control curves are bracketed to the range [0,1]. Consequently, if control values outside this range are
encountered it is immediately obvious that this is outside the capability of the curves. The second
improvement is the introduction of a simple model of the volumetric distribution of hull forms in order
to reduce interdependency between control parameters affecting the fullness of the hull forward and
aft and the goal parameters of volume displacement and longitudinal centre of buoyancy (LCB). A
simplistic block barge model of a section area curve is used to represent the fixed volume (analogous
to the volume in the parallel middle body) and the variable volume in the entrance and run of the
hull, Fig.19. The equivalent mathematical model is:
T = CF F + M + CA A

(1)

T xT = CF F xF + M xM + CA A xA

(2)

denotes the displacement, X the longitudinal centre of buoyancy (LCB), and subscripts T , F , M ,
A, refer to the Total, Forebody, Midship and Afterbody volumes, respectively.

Fig.19: Simplified section area curve model used to represent fixed and variable hull characteristics
The model can be rearranged so that the value of the control parameters can be found in terms of the
target displacement and LCB:
CF

CA =

T XT M XM XA (T M )
F (XF XA )
T CF F M
A

(3)
(4)

The coefficients of the model are determined by exploring the minimum and maximum limits of
the control variables. The iterative process subsequently determines the exact values of the control
parameters by considering the error between the simple model and calculated hydrostatics of the hull
surface. This technique has been found to very successful, capable of working with hull forms without
parallel middle body (e.g. YachtLINES) and is easily adapted to finding partial solutions when one of
the target parameters cannot be achieved.
8.4

Local Surface Modification

The details of the technique described so far develop a hull surface by blending between the different
transverse definition curves. This process depends on the definition of the longitudinal blending curves
which must clearly define a family of shapes for the hull form to be fair. This framework does not
allow for local features such as appendages to be introduced, as this prevents the longitudinal blending
curves from being controlled as a family.
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

131

Hull surface manipulation is an iterative process and it is often difficult to see whether a single
change makes a positive or negative step towards producing a fair form. The manipulation process is
made somewhat more difficult by the standard stack approach used to implement undo/redo because
the user may have to undo desirable changes to reach the undo information for earlier undesirable
changes made elsewhere in the surface. As a result, there has been little impetus to introduce additional
modelling tools that assists by modifying larger portions of the surface because, although the operation
itself can be directly undone, in relation to the other changes that the user may subsequently make,
it may be detrimental to the hull surface design process as a whole.
This problem does not exist in IntelliHull as the hull generation process consists of a series of
operations that are reversible if necessary. Once an initial hull surface has been generated, tools that
make local modifications to the surface can be applied independently introducing features, such as a
bulb and other appendages, without affecting any preceding definition information. (The hull surface
is generated from the transverse definition curves which would not be affected by any subsequent
modelling operations applied directly to the hull representation).
Piegl and Tiller (1995) present several techniques for introducing locally shaped features into a
B-Spline curve or surface. These techniques operate by refining the curve or surface definition in
the area where the feature will be introduced by adding knots. Subsequently, the control points are
blended between their original location and the introduced shape. IntelliHull implements a variant of
the warping process to introduce a parametric bulb into the hull surface. Rather than modify the hull
representation by blending between the two shapes, as in Piegl and Tiller (1995), the definition of
the hull surface is refined allowing the control vertices from the hull surface to be moved to the same
location as control points defining the bulb surface. The details of the operation are as follows:
1. The bulb surface definition is orientated so that U/V directions are compatible with the main
hull surface.
2. Bulb surface control vertices which lie outside of the main hull surface are identified establishing
a selection of vertices to be introduced into the main hull surface definition, Fig.20a.
3. Hull surface control vertices lying inside of the bulb surface control polygon are identified establishing the equivalent selection of vertices that will be moved onto the bulb surface.
4. Knots are introduced into either surface until the selection of vertices on each surface has the
same number of vertices, Fig.20b.
5. The selected control vertices of the main hull surface definition are moved to the location of the
equivalent selected control vertex in the bulb surface.
6. As the bulb surface is defined by discrete points, unlike the continuous mathematical functions
used to modify surface shape in Piegl and Tillers examples, blending must be explicitly applied
around the boundary of the modified area.

a) Identify control ver- b) Refine the hull con- c) Move control points d) Resulting Hull Contices outside the hull
trol polygon and iden- and fair at the intersec- tours
tify points to move
tion
Fig.20: Procedure to introduce a bulb into a hull surface control polygon
132

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

Although the shape of main surface does not exactly match the originating surface of the bulb
feature, the result is more than accurate enough for surface generation. This technique can be used
for introducing other appendages into the surface although this has not been investigated. However,
potential problems have been identified if the grid orientation of the control polygon of each surface is
not closely orthogonal to each other. In these cases, it may be better to follow the examples presented
by Piegl and Tiller (1995) more closely.
9

Examples of IntelliHull Definitions

9.1

Implementation

The IntelliHull technique is implemented in PolyCAD, a tool that provides a range of entities from
pure curve and surface definitions such as B-Splines to the proprietary hull generation techniques
including YachtLINES and ShipLINES. The user interface is optimised for hull surface design and
aims to provide all the basic tools a designer expects such as control point manipulation and hydrostatic calculations as well specific features such as allowing the user to actively control one and two
dimensional manipulation while viewing in three dimensions, snapping to other entities and analyse
curvature of both curves and surfaces. The PolyCAD implementation of IntelliHull takes advantage
of the editing features available to all interactive entities provided within the system. Two IntelliHull
entities are exposed to the user, the IntelliCurve and the IntelliHull itself.

Sketch initial midship section

curve

Use constraints to form correct shape

Add remaining curves at bow Define hull surface

and stern
Fig.21: Process to develop hull form
9.2

Add curves on parallel middle

body

Add bulbous bow

Generating a New Hull

The following section demonstrates some of the techniques capabilities for creating hull surface
definitions. The first example shows the main steps used to build up a new hull surface. The user
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

133

starts by sketching a curve, then applies two straight and one blend constraint to form a generic ship
type midship section curve. The midship section curve is twice copied to form the forward and aft
limits of the parallel middle body. The constraints are removed and an extrusion (offset) constraint
applied so that both curves have the same shape in section as the midship curve. Further copies are
made to form the basis for definition curves at the bow and stern. The bow tangent curve is defined in
a similar way to the parallel middle body, extruded in Y from the bow curve to ensure that the hull
surface is perpendicular to the centre plane along the stem. The hull surface is formed by blending it
through the shape of each curve. Additional attribute information is used to constrain the surface by
imposing specific tangent effects at the divisions of the parallel middle body and at the bow. Finally, a
parametrically defined bulbous is added by applying the local surface modification technique described
previously, Fig.21.
9.3

Using IntelliHull to Copy Surfaces

An ability to create surfaces based on existing hull forms is an essential feature. Fig.22 compares
a surface generated from the bow of a Ro-Ro hull form omitting the shape of the bulb. The resulting
surface is a fairly close match and differs along the stem where the original hull has more strength
in the tangent and along the forward flat of bottom curve. A comparison of hydrostatics shows that
the generated hull form was generally within 0.3% of displacement and within 2% of LCB. Fig.23
compares the aft end of a bulk carrier hull with a generated surface. In this example, the surface has
more difficulty in matching the original surface as the demonstrator does not have sufficient control
of the surface and requires additional trimming tools to produce the transom. Correspondingly, the
hydrostatics calculations indicate that the generated hull is within 10% of displacement and an error in
LCB of 30%. The second example illustrates that the single surface demonstrator has limitations for
hulls with local areas of high curvature which are not part of a specific feature or appendage. Despite
this, the resulting surface is reasonably fair and could be used as part of a new concept design.

Fig.22: Matching to bow of Ro-Ro hull

except the bulb
IntelliHull contours are dashed

10

Fig.23: Matching to the stern post of bulk

carrier; stern needs to be
trimmed to form transom.
IntelliHull contours are dashed

IntelliHull in an Integrated Ship Design Tool

While many of the parametric hull generation techniques discussed here have limitations, even the
simplest techniques can be used to generate good hull forms for particular design scenarios. However,
one of the main failings is that these techniques are generally pushed to the fringes of hull form design
by having very little connectivity to other tools used in ship design process. They are, e.g., implemented
as separate pieces of software within suites of larger design tools or implemented as special features
hidden away behind the larger features of integrated design tools. Consequently, it is difficult for

134

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

the designer to immediately see the potential benefits that can be obtained when these techniques are
introduced into the design process and prefer to remain with the tried and tested technique of manually
manipulating the surface definition. IntelliHull is implemented within Paramarine, an integrated ship
and submarine design tool which supports parametric connectivity between all parts of the tool. The
technique is now being actively used for developing hull forms for conceptual and exploratory ship
design.
10.1

Paramarine

Paramarine is an integrated ship and submarine design environment developed by embracing the
full capabilities of modern object-orientated software development. The tool itself features an objectorientated design framework which allows the parametric connection of all aspects of both the product
model and analysis elements together. The system supports analysis disciplines in common to most
ship design tools such as stability, powering and structural analysis, which when combined with
parametric connectivity, allow designers to build up complex designs using all of the features of the
solid modelling kernel provided by the industry standard Parasolid tool set. In addition, Paramarine
features several unique modules specifically orientated towards the development of concept designs
where the role of the vessel may require the designer to explore innovative solutions.
Objects make up the individual definition elements of a design usually with a set of corresponding
attributes which may be numerical values, other objects or references. The design itself will consist of
a collection of simpler objects assembled together to form a more complicated design. Therefore, the
topological aspects are modelled as accurately as the geometric aspects.
One of the key objectives of Paramarine is the management of the design configuration. In essence,
the system will attempt to maintain every aspect of the design in a correct and updated state by
responding to user changes. This not only means that out of date elements are recalculated but also
that any connections (references) are legitimate, e.g., ensuring that a connection points to the right
kind of object (type-checking) or that the dimensional units of a formula equate correctly taking into
account the input parameters and the expected outcome. With fairly limited experience any user can
quickly define and arrange a set of objects which can be used to investigate various characteristics of
a design.
10.2

Paramarine Early Stage Design Module

The Early Stage Design provides the designer with an alternative method of exploring ideas without being forced to follow the traditional approach of defining hull form, subdivision etc. Based on
the University College London (UCL) Functional Building Block methodology, Andrews and Pawling
(2003), the framework consists of a hierarchy of objects called building blocks used to represent the
different functional aspects of the design. The module provides the designer with a free space to
construct solutions to requirements and determine the overall form of the vessel. While the methodology can be considered abstract when compared with the traditional approach, the process is much
more rigorous as it forces the designer to define exactly how the solution will function. Consequently,
decisions are addressed earlier reducing the amount of time it takes to deliver a design.
Building blocks in the design are associated with attributes which encapsulate the geometric, functional and topological characteristics to clearly define the role of a block. Characteristics (Weight,
Space, Buoyancy, Consumable, Manning, etc.) can be used to define both requirements and solutions
in the style of supply or demand. In many cases a block providing a solution to a particular problem may introduce further requirements which must be subsequently addressed. By working through
this design process all of the requirements associated with a design can be captured and solutions to
requirements defined. The Early Stage Design module collates information from within the building
blocks hierarchy and compares related characteristics together, a simple example being a comparison
between the buoyancy and weight characteristics to see if the design will float. Ultimately, the designer aims to develop a balanced design by making decisions based on information extracted from the
building block hierarchy. The geometrical definition of individual building blocks is initially coarse to
Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

135

minimise the level of detail. The design is likely to go through several stages of refinement starting
with a coarse definition to allow the primary design decisions to be evaluated moving to finer models,
as the details of the solution are worked through, Fig.24.

Fig.24: The impact different design solutions can be flexibly investigated e.g. machinery arrangement
(left) and (centre) before adding more detail to the design for deeper analysis (right).
11

Continuing Work

While IntelliHull successfully demonstrates the concepts behind the design tool, the implementation exhibits limitations as it is applied to detailed design scenarios. Two key limitations are that
the transverse control curves provide the designer with very limited control over longitudinal shapes
particularly for fine hull forms and the application of successive local surface modification creates
a surface which, while accurate and fair, has too many control point and non-uniform knot vectors
making it difficult to modify the definition in any other tool.
Surfaces defined by multiple patches offer a much more suitable platform for this technique as local
features can be defined in greater resolution independently of other parts of the surface. Furthermore,
as these techniques are usually driven by a grid of curves, hull form topology can be defined more
clearly than in the demonstrator allowing the transformation functions to be implemented as implicit
features of the definition. Based on experience using IntelliHull, the curves will not directly expose
the user to the underlying NURBS representation but will instead construct the control polygon based
on the users point data and constraints allowing straight sections, knuckles, blends or cubic splines to
be again produced. The surface itself will be generated by forming patches between each face of the
manifold network of curves using subdivision to accurately extract each edge as a NURBS curve.
Work to develop this solution has been progressing steadily for some time and the greatest proportion involves the development of a technique which can create surfaces from an irregular network
of curves, an operation which several existing tools already offer. At this time, despite being unable
to produce a complete surface for complex hull ship forms, simpler hull forms such as yachts can be
defined and some basic topological transformation capability can be demonstrated, Fig.25.

a) Curve network for yacht

constructed from linked BSpline curves

b) Surface resulting from (a)

c) Ship surface; holes are faces with

more than 4 sides, a configuration
not yet implemented

Fig.25: Surfaces produced by the IntelliHull technique

136

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

References
ANDREWS, D.; PAWLING, R. (2003), SURFCON A 21st century ship design tool, 8th Int. Marine Design
Conf. (IMDC), Athens
BENSON, F.W. (1940), Mathematical ships lines, Trans. RINA 82
BEZIER, P. (1972), Numerical control, mathematics and applications, John Wiley & Sons
BLOOR, M.I.G.; WILSON, M.J. (1990), Using partial differential equations to generate free-form surfaces,
Computer-Aided Design 22
BOLE, M. (2002), A hull surface definition solution based on form topology and geometric constraints, Ph.D.
Thesis, Univ. of Strathclyde, Glasgow
CHAPMAN, F.H. (1760), A treatise on ship-building
CREUTZ, G.; SCHUBERT, C. (1978), Interactive curve creation from form parameters by means of B splines,
Schiffstechnik 25
HARRIES, S. (1998), Parametric design and hydrodynamic optimization of ship hull forms, Ph.D. Thesis, TU
Berlin, Mensch & Buch Verlag, Berlin
JORDE, J.H. (1997), Mathematics of a body plan, The Naval Architect, January
KUIPER, G. (1970), Preliminary design of ship lines by mathematical methods, J. Ship Research 14
LEE, K.Y.; RHIM, J.H.; LEE, S.U.; CHO, D.Y.; CHOI, Y.B. (2001), Development of sophisticated hull form
CAD system EzHull based on non-manifold model and X- topology, 8th Int. Symp. Practical Design of Ships
and Other Floating Structures (PRADS), Shanghai
LACKENBY, H. (1950), On the systematic geometrical variation of ship forms, Trans. RINA 92
LETCHER, J.S.; SHOOK, D.M.; SHEPARD, S.G. (1995), Relational geometric synthesis: Part 1 framework,
Computer-Aided Design 27/11, pp.821-832
LOWE, T.W. (2005), Design of partial differential equation hull surfaces using genetic algorithms, Ship Technology Research 52, pp.124-133
MUNCHMEYER, F.C.; SCHUBERT, C.; NOWACKI. H. (1979), Interactive design of fair hull surfaces using
B splines, ICCAS III, North-Holland
PIEGL, L.; TILLER, W. (1995), The NURBS book, Springer
REED, A.M.; NOWACKI, H. (1974), Interactive creation of fair ship lines, J. Ship Research 18/2
RIESENFELD, R.F. (1972), Applications of B-spline approximation to geometric problems of computer aided
design, Ph.D thesis, Syracuse University
ROGERS, D.F. (2000), An introduction to NURBS: with historical perspective, Morgan Kaufmann
STANDERSKI, N.E. (1998), The generation and distortion of ship surfaces represented by global tensor product
surfaces, PhD thesis, TU ???.
TAYLOR, D. W. (1915), Calculations of ships forms and light thrown by model experiments upon resistance,
propulsion and rolling of ships, Int. Cong. Engineering, San Francisco
VON KERCZEK, C. (1969), The representation of ship hulls by conformal mapping functions, J. Ship Research
13/4

Schiffstechnik Bd.53 - 2006/Ship Technology Research Vol. 53 - 2006

137