Design of practical structures can involve difficulties when it is based only on analytical models. Experimental data are therefore often made use of for such cases. There have been many studies on identification of statistical and dynamical characteristics of structural systems, making use of experimental data. In this study, we propose a method to identify the characteristics of structures with proper accuracy by using a hierarchical neural network (HNN) that has excellent projection ability even for systems which have nonlinearities. HNN has been generally used to learn dynamic characteristics with minimum number of experimental data. We use HNN as the solver of an optimum design, that is to estimate the dynamic characteristics and sensitivities. The effectiveness of this method is demonstrated through several numerical examples and experiments. Then we utilize the derived sensitivities for gradient-based optimization techniques and to solve the optimum design problems. It is found that the proposed method is valid and effective for such optimum design problems.
|ジャーナル||Nippon Kikai Gakkai Ronbunshu, C Hen/Transactions of the Japan Society of Mechanical Engineers, Part C|
|出版ステータス||Published - 1995 11|
ASJC Scopus subject areas
- Mechanics of Materials
- Mechanical Engineering
- Industrial and Manufacturing Engineering