[Published online Journal of Computer Chemistry, Japan Vol.19, 167-168, by J-STAGE]
<Title:> 人工知能による片状黒鉛鋳鉄の黒鉛組織の判定
<Author(s):> 池原 瑞生, 内田 希, 岩見 祐貴, 加藤 雅也, 菅野 利猛
<Corresponding author E-Mill:> s183195(at)stn.nagaokaut.ac.jp
<Abstract:> The graphite structure of flake graphite cast iron is classified into 5 types of A,B,C,D,E or 6 types including chill. In usually, type A is good quality and type D is poor quality. As the proportion of D-type graphite increases, the tensile strength decreases.Inspection of graphite structure is one of the quality evaluations. However, there is no method for numerically determining the graphite structure of flake graphite cast iron. In this study, we created an AI that determines 6 types of graphite structure from A to chill using a Convolutional Neural Network (CNN).The created AI judged the graphite structure photographs that not used for learning with an average correct answer rate of 97.5%. In addition, when the graphite structure of one sample was analyzed by AI, it was confirmed numerically that the graphite structure deteriorated.
<Keywords:> Convolutional Neural Network, Flake graphite cast iron, Graphite structure
<URL:> https://www.jstage.jst.go.jp/article/jccj/19/4/19_2021-0017/_article/-char/ja/
<Title:> 人工知能による片状黒鉛鋳鉄の黒鉛組織の判定
<Author(s):> 池原 瑞生, 内田 希, 岩見 祐貴, 加藤 雅也, 菅野 利猛
<Corresponding author E-Mill:> s183195(at)stn.nagaokaut.ac.jp
<Abstract:> The graphite structure of flake graphite cast iron is classified into 5 types of A,B,C,D,E or 6 types including chill. In usually, type A is good quality and type D is poor quality. As the proportion of D-type graphite increases, the tensile strength decreases.Inspection of graphite structure is one of the quality evaluations. However, there is no method for numerically determining the graphite structure of flake graphite cast iron. In this study, we created an AI that determines 6 types of graphite structure from A to chill using a Convolutional Neural Network (CNN).The created AI judged the graphite structure photographs that not used for learning with an average correct answer rate of 97.5%. In addition, when the graphite structure of one sample was analyzed by AI, it was confirmed numerically that the graphite structure deteriorated.
<Keywords:> Convolutional Neural Network, Flake graphite cast iron, Graphite structure
<URL:> https://www.jstage.jst.go.jp/article/jccj/19/4/19_2021-0017/_article/-char/ja/
