Abstract
A structure optimal design using mathematical techniques, such as a Genetic Algorithm, needs evaluating some conditions of numerous design candidates. It is not efficient to use FEM for analyzing the strength of structures in spite of requiring accuracy in the evaluation. In this paper, The Neural Network System is built for evaluating structure with FEM solutions as teacher data. In order to improve approximation accuracy, leading functions are introduced to the input variables of a Neural Network System. Multiple linear regression method is also used for selecting design variables to be used. These techniques improve the accuracy of Neural Network System and reduce manpower used for structure optimal design. A container ship is optimized and its result is shown here.
