Enhancing the durability of concrete in severely cold regions: mix proportion optimization based on machine learning

• Main Authors: Chen, Hongyu, Cao, Yuan, Liu, Yang, Qin, Yawei, Xia, Linyue
• 其他作者: School of Civil and Environmental Engineering
• 格式: Article
• 語言: English
• 出版: 2024
• 主題: Engineering::Civil engineering; Concrete Durability; Mix Proportion Optimization
• 在線閱讀: https://hdl.handle.net/10356/172953
• 機構: Nanyang Technological University

Compared with inland areas, the environment for concrete in severely cold regions is more harsh and includes ion erosion, changes in dry and wet conditions, low-temperature freeze–thaw cycles and other occurrences leading to surface damage to and the cracking of roads and bridges, which causes concrete structures to be unable to reach their service life due to insufficient durability. By optimizing the mix proportion (MP) of the materials used, the frost resistance and impermeability of concrete can be improved to enhance its durability. In this paper, we develop an intelligent prediction and optimization model of concrete durability (CD) based on a random forest (RF) model and NSGA-II, and a Pareto front of the optimal trade-off solutions can be obtained by using NSGA-II to perform the optimization. A final optimal solution, the one that is nearest to the ideal solution, is determined as the suggestion for decision-making. The research is verified by taking a key highway engineering project in the Plan for Revitalizing Northeast China as an example, and the results show that the following: (1) The key factors after screening are the water–binder ratio, cement content, coarse aggregate content, fine aggregate content, high-efficiency water-reducing agent and fly ash content. (2) In comparison with other machine learning algorithms, a filtered RF prediction model has high precision, the goodness of fit (R2) of the frost resistance and impermeability is higher than 0.95, and the root mean square error (RMSE) is less than 0.1. (3) After optimization, the chloride ion permeability coefficient of concrete is reduced by 47.9%, the relative dynamic elastic modulus (RDEM) is increased by 4.07%, and the cost is reduced by 2.4%. In summary, the proposed RF-NSGA-II intelligent hybrid optimization algorithm can improve the durability of concrete in severely cold regions while realizing the economic and environmental protection production of concrete to improve engineering safety performance and service life.