Automated system for structural design using design window search approach: its application to fusion first wall design

Abstract

Some practical structures are required to be operated in severe environments where various mechanical, thermal or electromagnetic loadings occur in a complicated manner and various failure phenomena such as yielding, fracturing and melting compete among others. In designing such structures, it seems very useful to determine the design window (DW) that schematically indicates an area of satisfactory solutions in a permissible design space. The DW may give us more meaningful information than one satisfactory or optimum solution. However very little research has been performed on this topic so far. This paper describes a novel automated system for structural design based on the DW concept. The present system consists of three main modules and one sub-engine, that is, (a) the Analyzer, (b) the DW Search Engine, (c) the Design Modifier as the main modules and (d) a multilayer neural network as the sub-engine of the DW Search Engine. The Analyzer takes care of computational mechanics simulations for various sets of design parameters. To maintain high flexibility and extensibility, the Analyzer is constructed using an object-oriented knowledge representation and a data-flow processing technique. Some DW search methods are incorporated into the DW Search Engine. Using neural network technology, DWs are searched very efficiently. The Design Modifier has the role of finding one satisfactory solution, which is given to the DW Search Engine so that some search methods can start searching. To demonstrate the practical performance of the present system, it is applied to design one of the typical structures to be operated in a severe environment, i.e. the first wall structure of ITER (International Thermonuclear Experimental Reactor).