有限要素モデリングのための事例類似性の評価

概要

論文の詳細を見る
This paper deals with support system for finite element modeling and with a methodology for case base retrieving. As well-known FEA (Finite Element Analysis) programs, there are general purpose codes such as MARC, ABAQUS, NASTRAN, etc. It is not easy to design input data, because of various multifunctionalities. Thus, finite element modeling supported with CBR (Case Based Reasoning) is required practically. It is recognized that CBR does not require high level enormous empirical knowledge, but needs a few of typical and essential cases. On the other hand, since there are many complex attributes for input data modeling, adapting methods of similar cases and its indexing methods from case base should be defined clearly. In this paper, a method of adapting similar cases has been proposed. This method derives metric distance between similar cases for themselves as follows : 1) evaluation of stochastic amount of attributes in a case population, 2) vectorized coding of attributes in a case, and 3) evaluation of vector angle between similar cases. Through several but sufficient retrieval simulations for similarity, availability of entropy weighting is recognized, and applicability of noise filter for similarity is shown.
日本シミュレーション学会の論文
1995-09-15