EFFECT OF AERODYNAMIC NOISE FROM A WALL-CONTACT TYPE UAV ON THE TILE DETACHMENT DIAGNOSIS USING RANDOM FOREST-BASED MACHINE LEARNING

Abstract

The hammering test for diagnosing tile detachment on exterior walls faces issues such as high scaffolding costs and variability due to inspector differences. This study uses a wall-contact-type UAV with a vibration-based test hammer and applies Random Forest-based machine learning to diagnose detachment from the impact sound. The purpose of this study is to clarify the effect of UAV aerodynamic noise on feature extraction and diagnostic accuracy, and to improve performance via hyperparameter tuning. Results show that aerodynamic noise reduces the mean F1 Score, but tuning improves accuracy even under noisy conditions, supporting the method’s feasibility in practice.