機械学習を援用したクリープ・疲労損傷評価に関する検討

抄録

Recently, the development project of A-USC plants which have superior efficiency compared with USC plants have been underway from the requirement of the energy mix. Nickel-base heat resistant alloys are being considered as a candidate for their structural materials. In the A-USC plants, the structural materials used in the hot section undergo not only creep damage during the steady-state operation at elevated temperatures but also creep fatigue damage due to the loading transients including startup and shutdown. Therefore, creep-fatigue interaction becomes important as a major damage mechanism for structural materials in A-USC plants, however, the remaining life assessment technique have not been established considering the creep-fatigue damage. Semi-destructive methods are considered a useful inspection method for structural materials of high temperature plants because such techniques can be applied with the miniature-sized sample that can be cut from the actual component with smallest damage. Especially, EBSD analysis has advantages because the degree of deformation or damage can be expressed quantitatively as a local change in crystal orientation. It can be expected that machine learning utilized to mine data obtained from EBSD analysis characterize the damage in actual components. In this study, a machine learning method was investigated to analyze the damage mode and remaining life of Ni-based alloys subjected to creep, high-temperature fatigue, and creep-fatigue damages.