초음파 속도를 활용한 모르타르의 초기동해 예측 모델 제안 및 적용성 검토

초록

This study evaluated the applicability of a multi-stage regression model based on high-frequency ultrasonic testing and machine learning for diagnosing early frost damage in mortar. In the first stage, compressive strength was predicted using ultrasonic pulse velocity and curing age; in the second stage, the predicted compressive strength was used to estimate the depth of early frost damage. Among the four regression algorithms applied, the combination of the 250 kHz ultrasonic data and the gradient boosting model consistently showed the highest prediction accuracy. These findings suggest that the integration of high-frequency ultrasonic signals with nonlinear regression models is effective in assessing microstructural damage caused by early frost, and can serve as a foundation for future applications in concrete structures and the development of on-site diagnostic technologies.

키워드