|
Title |
Data-Driven Classification of Water-to-Cement Ratio Based on Electrochemical Impedance Features
|
|
Authors |
박주혜(Joohye Park) ; 이준영(Junyoung Lee) ; 홍진영(Jinyoung Hong) ; 최하진(Hajin Choi) |
|
DOI |
https://doi.org/10.4334/JKCI.2026.38.2.139 |
|
Keywords |
물-시멘트비; 임피던스 분광법; 시멘트 페이스트; 랜덤 포레스트 water-to-cement ratio (w/c); impedance spectroscopy; cement paste; random forest |
|
Abstract |
The water-to-cement ratio (w/c), a key indicator of concrete quality, has a significant influence on both compressive strength and durability. Theoretically, the minimum w/c required for complete hydration is 0.36. In practice, however, discrepancies between the design and actual values frequently arise due to factors such as mixing mistakes, evaporation loss, and residual wash water. This highlights the need for a simple and accurate on-site method to evaluate the w/c ratio. We propose a classification model based on electrical impedance data to rapidly identify concrete mixtures that fail to meet the minimum w/c threshold (36 %) during site acceptance. Unlike conventional equivalent circuit model (ECM) approaches, our proposed method utilizes raw impedance data from early-age cement paste. After statistically confirming its classification potential, a random forest algorithm was employed to build an optimized model. The model was evaluated using cement paste datasets with various w/c ratios, achieving a classification accuracy of 92.5 % and an area under the curve (AUC) of 0.982 for the 36 % threshold, along with a macro-average F1-score of 0.917. These results demonstrate the feasibility of detecting substandard concrete through simple measurements and analysis, and suggest its potential for integration into automated quality control systems.
|