Scholarlink

To achieve an optimal balance among concrete’s technical properties, environmental impacts, and economic costs, this chapter proposes a three-level ensemble learning framework for concrete strength prediction and then uses NSGA-II and TOPSIS method for multi-objective optimization of strength, cost, and carbon emissions. The results show that stacking and voting methods have better model performance in terms of predicting concrete strength. NSGA-II can effectively obtain Pareto solutions when strength is lower than 50 Mpa, which may be due to limited datasets. A larger concrete mix dataset is necessary to obtain robust optimization results. Future research can focus on incorporation of domain knowledge into machine learning model, a hybrid of MOO algorithms, more interactive Pareto pruning methods based on decision makers’ preferences, etc.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Low Carbon-Oriented Concrete Mix Optimization Using Ensemble Learning and NSGA-II

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年

查看全文
  • Lin DENG,
  • Xueqing Zhang

摘要

To achieve an optimal balance among concrete’s technical properties, environmental impacts, and economic costs, this chapter proposes a three-level ensemble learning framework for concrete strength prediction and then uses NSGA-II and TOPSIS method for multi-objective optimization of strength, cost, and carbon emissions. The results show that stacking and voting methods have better model performance in terms of predicting concrete strength. NSGA-II can effectively obtain Pareto solutions when strength is lower than 50 Mpa, which may be due to limited datasets. A larger concrete mix dataset is necessary to obtain robust optimization results. Future research can focus on incorporation of domain knowledge into machine learning model, a hybrid of MOO algorithms, more interactive Pareto pruning methods based on decision makers’ preferences, etc.