1834 ニューラルネットワークを用いた結合剛性推定法(G03-5 ニューラルネットワーク・非破壊検査,G03 材料力学)

概要

論文の詳細を見る
A rigidity of a part connected by thin-walled box section beams is one of the key factors dominating the basic characteristics of an automobile body structure. In the beginning stage of the body design, it is important to evaluate the rigidity of jointed part. In this study, for the Y-shape joint structure composed of three thin-walled beams, we propose the new technique that can quickly and accurately estimate the joint stiffness values from the design parameters (plate thickness, dimensions of a cross section, etc.). And the tool was made by using this technique. Main flow of this technique is as follows. At first, some learning data of the design parameters and the joint stiffness values is prepared. Next, the relation between the design parameters and the joint stiffness values is constructed by feedforward neural network. Once the relation between both was constructed, the joint stiffness values can be obtained quickly by inputting the design parameters to the trained neural network.
一般社団法人日本機械学会の論文
2005-09-18