転移学習を適用したRC梁の損傷判定におけるアンサンブル学習導入による判定性能向上

抄録

This study proposes a non-destructive method for damage assessment of reinforced concrete (RC) beams by combining transfer learning with an ensemble Local Outlier Factor (LOF) approach．A selective transfer strategy was introduced to address domain mismatch between source and target data, while Bagging－based ensemble LOF was applied to improve accuracy and stability under small－data conditions．The results indicate that integrating transfer learning with ensemble methods provides a robust and reliable framework for RC beam damage assessment．