設計偏差パターンの類似度解析による故障モードの判別分析の試み

抄録

This study aims at developing multi-class clustering method of damage/fracture modes using Support vector machine (SVM).Conventional failure mode analyses methods lack the method how the deviation in design specification associate with specific types of failure modes. In our proposed design deviation method, an association table between design/ environmental deviation with the types of damage/ fracture modes has been made. The association table can help practitioners to predict as many as possible failure modes in components caused by damage/ fracture modes of materials. Multi-class clustering using SVM demonstrated that the precision values of each damage/ fracture modes were 77 -100 %. The result demonstrated that the proposed multi-class clustering method can support in selecting specific damage / fracture mode.