Vlasenko Tetiana, Hutsol Taras, Vlasovets Vitaliy, Glowacki Szymon, Nurek Tomasz, Horetska Iryna, Kukharets Savelii, Firman Yuriy, Bilovod Olexandra
Department of Management, Business and Administration, State Biotechnology University, Alchevsky St., 44, Kharkiv, 61002, Ukraine.
Department of Mechanics and Agroecosystems Engineering, Polissia National University, Stary Boulevard 7, Zhytomyr, 10-008, Ukraine.
Sci Rep. 2025 Jan 7;15(1):1210. doi: 10.1038/s41598-024-84996-8.
The effective implementation of the European Green Deal is based on closing cycles by means of reusing products and extending their durability, especially for steel products in the construction industry. The Life Cycle Assessment gives an opportunity to determine the potential impact caused on the environment by building structures and it is used mainly at the early design stage. At the same time, there are significant gaps when it comes to predicting properties of steel products at the last stage of the life cycle of existing buildings in the End of Life Stage (C1-C4) phases and especially D-Benefits and Loads Beyond the System Boundary. This paper uses machine learning (ML) in order to solve the problem of predicting the reusability of construction steel based on the determination of its yield strength by a non-destructive magnetic method. This will give an opportunity to make informed decisions when using this steel again. The research uses machine learning approaches that include regression problems. However, the use of ensemble learning significantly improves quality and accuracy of the results, while demonstrating its advantage in combining multiple models for obtaining more accurate predictions. The research results show that the WeightedEnsemble ensemble method (which includes 8 models) has the best prediction accuracy (MSE = 441 MPa and RMSE = 21 MPa). This method has high accuracy and low delay of conclusion (IL = 0.119 s) when predicting the tensile strength limit (MPa) based on the data of non-destructive testing of structural steel products. . The innovation of the development lies in the ability to provide an automated tool to support informed decision-making for the reuse of building steel for construction site professionals. The accuracy of the ensemble model and its potential for integration with existing practices indicate significant progress in managing steel reuse processes at the final stage of the building life cycle - stage D.
欧洲绿色协议的有效实施基于通过产品再利用和延长其耐用性来实现循环闭合，特别是对于建筑业中的钢铁产品。生命周期评估为确定建筑结构对环境造成的潜在影响提供了机会，并且主要用于早期设计阶段。同时，在预测现有建筑物生命周期最后阶段（报废阶段C1 - C4），尤其是系统边界之外的D - 效益和负荷时，对于钢铁产品性能的预测存在显著差距。本文使用机器学习（ML）来解决基于通过无损磁方法确定建筑用钢的屈服强度来预测其可再利用性的问题。这将为再次使用这种钢材时做出明智决策提供机会。该研究使用了包括回归问题在内的机器学习方法。然而，集成学习的使用显著提高了结果的质量和准确性，同时展示了其在组合多个模型以获得更准确预测方面的优势。研究结果表明，加权集成（WeightedEnsemble）集成方法（包括8个模型）具有最佳预测精度（MSE = 441 MPa，RMSE = 21 MPa）。当基于建筑用钢产品无损检测数据预测抗拉强度极限（MPa）时，该方法具有高精度和低结论延迟（IL = 0.119 s）。该开发的创新之处在于能够为施工现场专业人员提供一种自动化工具，以支持对建筑用钢再利用做出明智决策。集成模型的准确性及其与现有实践集成的潜力表明在建筑物生命周期的最后阶段 - D阶段管理钢材再利用过程方面取得了重大进展。
