ステンレス配管の肉厚評価に有効な特徴量の検討

抄録

In this study, experimental modal analysis system for extracting features used for machine learning has been constructed by combining experimental setup and MATLAB programming. The experimental setup could monitor piping vibration produced by hammering. The MATLAB programming could automatically extract five features from the vibration data obtained by the experimental setup. The constructed experimental modal analysis system was applied to stainless steel pipes with different wall thickness, and the feature importance was discussed by performing machine learning. It was found that the natural frequency was promising feature for the evaluation of the pipe wall thinning.