Title:
Prediction of Relative Dynamic Elastic Modulus of Manufactured Sand Concrete Based on Machine Learning (Prepublished)
Author(s):
Jinpeng Dai, Jieyu Zhou, Yu Chen, Lei Li, Xuwei Dong
Publication:
Materials Journal
Volume:
Issue:
Appears on pages(s):
Keywords:
fineness modulus; machine learning; manufactured sand concrete; parent rock lithology; relative dynamic elastic modulus
DOI:
10.14359/51749413
Date:
12/18/2025
Abstract:
The durability of manufactured sand concrete is substantially influenced by variations in parent rock lithology, fineness modulus, and stone powder content of the manufactured sand. This study develops a predictive model for the relative dynamic elastic modulus of manufactured sand concrete using six machine learning algorithms. The results demonstrate that the CPO (crested porcupine optimizer)-optimized XGBoost model exhibits superior prediction accuracy and stability. The algorithm-based optimization reveals that manufactured sand produced from limestone, iron ore tailings, and quartzite demonstrates improved frost resistance in concrete. The optimal fineness modulus range was found to be 2.6 to 2.86; stone powder content should be maintained between 3 and 12% for optimal performance. The study further proposes a mixture ratio optimization scheme that takes into account frost resistance, material cost, and carbon emissions, so that the cost and carbon emissions of single concrete are reduced, and the frost resistance is further improved.