Page 311 - Practical Design Ships and Floating Structures

286

Design is by nature an evolving process, and incremental improvements are made by designers based
on past experience and knowledge gained in executing this process. Previously tried and tested designs
are often re-used and improved upon to anive at even better designs. This form of design reuse is
routinely used in many areas of marine design. For example, many ships are designed from basis ships
which are known to possess good performance characteristics. However, the complexity of a marine
design problem does not always allow efficient application of design reuse paradigms. Decomposition
breaks down complexity of design problems and hence allows designers to reuse design data in a
relatively efticient manner.

This paper discusses some aspects of the application of a decomposition and reuse approach in marine
design. The decomposition strategy employed allows identification of weakly connected model sub-
structures that naturally exist within a design problem and attempts to divide the design problem into
sub-problems in accordance with these sub-structures. Reuse of existing design data to solve the
decomposed design problems further improves the efficiency and quality of the design solutions. A
ship design example is used to illustrate the application of this decomposition and reuse approach in
marine design.

2 DESIGN DECOMPOSITION

The motivations to decompose a design problem into a number of smaller sub-problems are: reduction
of problem complexity, application of different targeted solution procedures on different sub-problems,
carrying out problem-solving activities concurrently and utilising parallel computing opportunities.

While decomposition is a widely used problem-solving approach, there are significant variations in the
criteria and strategies used for performing design decomposition in practice. Design decomposition
strategies based on a combination of structures (physical components, logical objects, etc.), behaviours
(action, force, process), disciplines, and goals or fimctional requirements (design properties that satisfy
given requirements) have been observed (Koopman 1995). For designs that involve sequential flow of
information sequential decomposition may be appropriate (Scott and Sen 1998). In addition, various
methods for dealing with design sub-spaces problems exist, for example in the area of optimisation in
design subspace based on decomposition (Rao and Sen 1999).
When a design problem is broken down into a number of simpler sub-problems, the problem-solving
procedures then involves solving such sub-problems of reduced complexity. It is desirable, therefore,
to reduce the interactions between sub-problems (i.e. to reduce the co-ordination effort). However,
many decomposition approaches do not take the magnitude of this co-ordination effort directly into
consideration so that the decomposed sub-problems may have a relatively large number of variables
that are common between the decomposed sub-problems. Put simply, the main problem may be
decomposed into a number of highly coupled sub-problems requiring a relatively large co-ordination
effort during the problem solving process. The additional co-ordination effort required in solving a
series of highly coupled sub-problems might significantly undermine the initial objective of breaking
down complexity by decomposition.
A hypergraph based approach to design decomposition is used in this paper. In this approach, a design
problem is represented by a hypergraph. A hypergraph consists of nodes connected by hyperedges. A
hyperedge connects two or more nodes. The objective of the hypergraph partitioning operation is to
decompose a given hypergraph into a desired number of partitions (sub-hypergraphs) such that the
hyperedges spanning across two or more partitions are minimised. The hypergraph partitioning
approach has been applied in decomposition of Very Large Scale Integration (VLSI) circuit design
problems, cell formation problems in flexible manufacturing systems and Finite Element Method
(FEM) problems, to name but a few.