기계학습에 기반한 AA1100 합금의 부식에 미치는 환경 영향 분석
Machine Learning-Based Analysis of Environmental Effects on the Corrosion of AA1100 Alloy
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study explores the effects of three environmental chemical species—Na2S, NaCl, and H2O2 —on the corrosion behavior of AA1100 aluminum alloy using machine learning (ML) techniques. We collected experimental data through a full factorial design that included 27 solution conditions, from which we extracted two key electrochemical parameters: corrosion potential (Ecorr) and corrosion current density (icorr), We trained four regression algorithms—k-nearest neighbor, random forest, support vector regression, and extreme gradient boosting—on this data and compared their performance using root mean squared error. Among these models, the random forest algorithm demonstrated the most accurate predictions for both Ecorr and icorr, leading us to select it for further analysis. To assess the influence of each input variable, we employed SHAP (SHapley Additive exPlanations) analysis. Our findings revealed that NaClmade the greatest positive contribution to Ecorr, while H2O2 significantly increased icorr. In contrast, Na S2 had minimal impact on both corrosion indicators. SHAP interaction plots indicated that Cl⁻ and H2O2 had synergistic effects in accelerating corrosion, while Na2S remained inert. This study highlights the effec- tiveness of interpretable ML approaches in analyzing complex corrosion systems and offers a data-driven perspective for corrosion prediction and material design.
Abstract
This study explores the effects of three environmental chemical species—Na2S, NaCl, and H2O2 —on the corrosion behavior of AA1100 aluminum alloy using machine learning (ML) techniques. We collected experimental data through a full factorial design that included 27 solution conditions, from which we extracted two key electrochemical parameters: corrosion potential (Ecorr) and corrosion current density (icorr), We trained four regression algorithms—k-nearest neighbor, random forest, support vector regression, and extreme gradient boosting—on this data and compared their performance using root mean squared error. Among these models, the random forest algorithm demonstrated the most accurate predictions for both Ecorr and icorr, leading us to select it for further analysis. To assess the influence of each input variable, we employed SHAP (SHapley Additive exPlanations) analysis. Our findings revealed that NaClmade the greatest positive contribution to Ecorr, while H2O2 significantly increased icorr. In contrast, Na S2 had minimal impact on both corrosion indicators. SHAP interaction plots indicated that Cl⁻ and H2O2 had synergistic effects in accelerating corrosion, while Na2S remained inert. This study highlights the effec- tiveness of interpretable ML approaches in analyzing complex corrosion systems and offers a data-driven perspective for corrosion prediction and material design.