2014 年 80 巻 810 号 p. SMM0020
This paper is about damage diagnosis using discriminant analysis via the support vector machine(SVM). For evaluation of reliability of the structure, condition of the structure is divided to several condition levels by the threshold in many cases. At these cases, identification of the damage as the discriminant levels is sufficient for the damage identification model. This paper is focused on the improvement of the identification accuracy of damage identification using discriminant analysis. SVM is one of linear learning machine which conduct 2 groups discrimination which used for discriminant analysis and pattern recognition. SVM is enable to use for the analysis which the boundary of the groups has nonlinear relation. About the damage identification, measurement data is affected by location and stage of the damage and simple linear discrimination is difficult to adopt as the method. In this paper, SVM is applied to identification of a delamination cracks in CFRP beams using the electric resistance change method. The effectiveness of this method is investigated from FEM model of electric resistance change due to the growth of the delamination crack. As a result, with weighted method for avoiding the risk side evaluation, the risk side evaluation which cause degradation of the reliability of structure was reduced.
