AUTHOR=He Zhengfeng , Wu Zhuofan , Niu Wenjun , Wang Fengcai , Zhong Shunjie , Han Zeyu , Zhao Qingxin 

TITLE=A machine learning model for predicting the mechanical strength of cement-based materials filled with waste rubber modified by PVA

JOURNAL=Frontiers in Materials

VOLUME=Volume 11 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/materials/articles/10.3389/fmats.2024.1490006

DOI=10.3389/fmats.2024.1490006

ISSN=2296-8016

ABSTRACT=To further expand the application of waste rubber in civil engineering, this study proposes a method for modifying waste rubber with polyvinyl alcohol (PVA). A machine learning database was established based on the mechanical strength of cement-based materials filled with PVA-modified waste rubber. Various machine learning methods were employed to develop regression prediction models, compare their accuracy, and analyze the robustness of different influencing factors on strength indicators. Specifically, the Support Vector Regression (SVR) model demonstrated superior prediction performance, with a mean squared error (MSE) of 1.21 and 0.33, and a mean absolute error (MAE) of 2.06 and 0.15. The analysis revealed that rubber content and w/c ratio were negatively correlated with strength indicators, while curing age and PVA showed a positive correlation. Among all influencing factors, rubber content had the most significant impact on strength. Additionally, the results indicated that PVA-modified waste rubber significantly improved the mechanical strength of cement-based materials, likely due to the dispersing and bridging effects of PVA. This method not only enhanced material performance but also helped reduce the environmental burden of waste rubber, offering significant economic and environmental benefits.