Page 78 - Practical Design Ships and Floating Structures

53

For smaller units (sailboats, small craft, etc.), the cylindrical zone is smaller, or even non-existent. In
this case, the LBR-5 model can be used to perform transverse cross-section optimization (midship
section).
The module can also be used in the final stage of the project to perform a general verification or to
refine the scantling. In addition, LBRS can be advantageously used for education and training purposes,
for instance to support lectures on ‘Ship Design Methodology’, ‘Structure Analysis’, ‘Ship
Optimization’, etc. Many papers and books have been written on design philosophy and methodology,
both present and future. The most well known methodology for the design of naval and marine
structures is the “Design Spiral”. Despite its age, it is still used. However the current tendency is to
break with this design process and move towards “Concurrent Engineering”. A comprehensive
bibliography review related to design methodology is presented in Rig0 (2001.c).
LBR-4 (Rigo 1992), the previous version of the “stiffened panel method” for elastic analysis of
stiffened structures, was the starting point for the development of the LBR-5 optimization module
presented in this paper. The role of LBR-4 is to provide a fast and reliable assessment of the stress
pattern existing in the 3D stiffened structure.

The LBRJ software is the result of the integration inside the Same package of the LBR-4 (Rigo 1992)
and CONLIN (Fleury 1988) software and constitutes a new tool to achieve scantling optimization of
midship section. Methods similar to LBR-5 are proposed by, for instance, Hughes and a1 (1 992) and
Rahman and a1 (1995). LBR-5 is essentially preliminary design oriented. The structure modelling is
simple and fast, but not over-simplified.

The optimized scantling can be obtained within a couple of hours (maximum 1 day for complex
structures if starting from scratch). LBR-5 does not have the capability of a finite element analysis and
is restricted to prismatic structures and linear 3D analysis. But, on the other hand, LBR-5 uses explicit
exact first order sensitivities (derivatives of the constraint and objective functions by the hundreds of
design variables). Heavy and time consuming numerical procedures are not required. Sensitivities are
directly available as the method is based on an analytic solution of the differential equations of
cylindrical stiffened plates using Fourier series expansions. So., sensitivity formulations are known
analytically. In addition LBR-5 does not need to use the concept of local and global design variables.
Due to the efficient CONLIN mathematical optimization algorithm (convex linearization and dual
approach), optimization of the full structure can be performed with hundreds of design variables and
constraints using less than 10-15 global structure re-analysis.

2 LBR-5 AND THE CONCEPT OF “MODULE-ORIENTED OPTIMIZATION”

A multi-purpose optimization model, open to users and compatible with different codes and
regulations must contain various analysis methods for strength assessment that could be easily
enriched and complemented by users. The user must be able to modify constraints and add
complementary limitationdimpositions according to the structure type studied (hydraulic, naval,
offshore structures, etc), the code or the regulation in force and to his experience and ability in design
analysis. The objective is to create a user-oriented optimization technique, in permanent evolution, Le.
that evolves with the user and his individual needs. We define this as “Module-Oriented Optimization”.
The LBR-5 optimization model is based on this new concept and is composed of several modules.
Neither the module number nor their type is imposed. At the start, the whole model is made up of 3
basic modules (Fig. 1) and forms the framework of the tool (COST, CONSTRAINT and OPTI).
Around the COST and CONSTRAINT modules there are a large number of sub-modules. Each of
these sub-modules is specific to a type of constraint. In principle, it is necessary to have at least one