Page 339 - Practical Design Ships and Floating Structures

3 14

4 REASONING POLICY BASED ON LAYOUT DESIGN CASE

Naval ship compartment layout design reasoning policy based on CBR are showed as following.
(1) Firstly, to study and select the properties of compartment layout design problem, determining the
compartment layout design objectives, initial conditions, and restrictions.
(2) By using the case-indexing model of compartment layout design based on neural network, to select
a group of cases whose objective is like the current layout design and sort them by their degree of
similitude. Search the most similar case to current problem in the compartment layout design case
library and store it in the current workspace.
(3) Take the selected compartment layout design case as prototype of the solution, adjust and modify it,
complete the conversion of case, until it satises the compartment layout design requirement. The
conversion of case is realized by using compartment layout design knowledge inherited from layout
design cases. The knowledge is derived from design class or design prototype.
(4) If the prototype cannot meet the requirements even after modifications and adjustments, it must be
redesign. And the result of redesigning should store in cases in the new case form.

5 CONCLUSIONS
Using case presentation method and case-indexing policy based on neural network introduced, we can
realize the intelligent ship compartment 3d-layout design based on CBR.
The neural network model introduced in this paper have two major advantages when compared with
other case-indexing methods: 1) Neural network has ability of finding the hidden information in
samples. Via training, neural network construct a model. The experience of design expert using to
determine whether two ship compartment layout design cases are similar is hard to express by regular
form, while using neural network can refine the experience of experts into the network. It’s just the
problem general methods are hard to solve. 2) Neural network has the ability of self-adaptation.
Through the constant study, neural network can constantly modify the model to match the changing
exterior conditions. The study of new compartment layout design case samples make neural network
modify the weight matrix continuously. In this way, the experience of determiniig whether two ship
compartment layout design cases are similar used by design expert can be expressed more accurately.

References
[ 11MAO @an.( 1995). Research on Case Prototype Based Design Methodology and Design Support
System, A Dissertation Submitted to Huazhong University of Science and Technology for the Degree
of Doctor of Philosophy in Engineering
[2]Li Jun-Hua.(2000).Principles and Application on Intelligent 3D-Layout Design for Ship
Compartments Based Compound Knowledge Model, A Dissertation Submitted to Wuhan
Transportation University for the Degree of Doctor of Philosophy in Engineering