STUDY OF A REASONING SYSTEM FOR AVERAGE BUILDING WEIGHT BY NEURAL NETWORK

Abstract

This report proposes a reasoning system by neural network for average building weight that is needed for assuming member sections in structural planning, by using simple information such as structural material types and shapes. 6 input parameters are used on the basis of the first author's structural design experiences. The training and verification data are used from the works that were really designed or planned in FUJITA CORPORATION. From the result of the verification, this system is proved to be sufficient and effective in structural planning, though the data has to be carefully selected.