상세 보기
초록
This study developed machine learning models that predict the hardened properties of concrete using both mix proportion factors and fresh concrete properties as the input variables. For normal-strength concrete, the 28-day compressive strength was predicted from mix proportion factors and plastic viscosity, and the carbonation velocity coefficient was predicted from mix proportion factors, rheological parameters, and compressive strength. Correlation analysis showed moderate negative correlations of −0.43 to −0.49 between the rheological parameters and the carbonation velocity coefficient; however, since the carbonation velocity coefficient was calculated from the W/C using the Kishitani equation, this correlation may be an indirect relationship mediated by the W/C. The ANN model showed the best compressive strength prediction with R² of 0.9216 and MAPE of 0.0855, while the SVM model showed the best carbonation prediction with R² of 0.9112 and MAPE of 0.0940, both satisfying the generalization criteria. A comparison with models using only the mix proportion factors showed that adding the rheological parameters did not improve the prediction performance.
키워드
- 제목
- 배합요소 및 레올로지 정수를 활용한 머신러닝 알고리즘 기반 경화 콘크리트 물성 예측 모델 개발
- 제목 (타언어)
- Development of Machine Learning-Based Prediction Models for Hardened Concrete Properties Using Mix Proportion Factors and Rheological Parameters
- 저자
- 조찬영; 이유정; 김광원; 남경은; 한동엽
- 발행일
- 2026-09
- 유형
- Y
- 저널명
- 한국건설순환자원학회 논문집
- 권
- 14
- 호
- 3
- 페이지
- 456 ~ 469
- 언어
- KOR
- 출판사
- 한국건설순환자원학회
- 발행국가
- 대한민국
- 분량
- 14 페이지
- ISSN
- E 2288-369X
P 2288-3320