基于机器学习算法的机制砂混凝土配合比设计

Mix Design of Manufactured Sand Concrete Based on Machine Learning Algorithm

摘要:

采用5种机器学习算法来预测机制砂混凝土的工作性能和力学性能，提出了利用梯度提升决策树（GBDT）和XGBoost算法进行配合比设计的策略。在满足性能要求的前提下，通过误差分析迭代优化给出了机制砂混凝土的推荐配合比。结果表明，采用机器学习推荐配合比制备的机制砂混凝土，试验结果与设计目标的误差在10%以内，验证了在特定性能要求下进行机制砂混凝土配合比设计的可行性。

Abstract:

Five machine learning algorithms were employed to predict the workability and mechanical properties of manufactured sand concrete and a strategy for mix design was proposed using gradient boosting decision tree（GBDT） and XGBoost. Under specific performance requirements， iterative optimization through error analysis yields recommended mix for manufactured sand concrete. The results show that the for manufactured sand concrete produced using the machine-learning-recommended mix has an error of within 10% of the target values， confirming the feasibility of mix design for manufactured sand concrete under specific performance requirements.