Details

Autor(en) / Beteiligte

• Wang, Shiqi
• Xia, Peng
• Gong, Fuyuan
• Zeng, Qiang
• Chen, Keyu
• Zhao, Yuxi

Titel

Multi objective optimization of recycled aggregate concrete based on explainable machine learning

Ist Teil von

• Journal of cleaner production, 2024-03, Vol.445, p.141045, Article 141045

Ort / Verlag

Elsevier Ltd

Erscheinungsjahr

2024

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Recycled aggregate concrete (RAC) has received extensive attention as a green building material. However, the service performance and sustainable potential of RAC have not been fully evaluated by traditional models due to the multi-component coupling mechanism. This study proposed a procedure to optimize the mixture of RAC using machine learning (ML) and multi-objective optimization (MOO) model, which enhances the compressive strength and chloride ion resistance of RAC while minimizing the effect on environmental impact (EI) and life cycle costs (LCC). A database containing 807 experimental samples was established to compare the prediction performance of ML models and swarm intelligence (SI) algorithms. WOA-BPNN has the highest prediction accuracy in predicting RAC compressive strength and electric charge passed, with R2 are 0.9904 and 0.9837 on testing set, respectively. The analysis of the feature importance by SHapley Additive explanation (SHAP) and partial dependence plot (PDP) indicates that cement, water and curing age are the key parameters affecting the compressive strength and electric charge passed. The MOO model established by setting constraints and objective functions can effectively evaluate the trade-offs between compressive strength, electric charge passed, EI and LCC, and all pareto solution sets for bi-objective and tetra-objective provide the optimal mixture scheme. Fly ash is the best supplementary cementitious materials (SCMs), as it can enhance the compressive strength and chloride ion resistance of RAC without increasing EI and LCC. The proposed optimization procedure can improve the efficiency of RAC mixture proportion considering mechanical, durability, EI and LCC.