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

Abstract

As demand for sustainable building materials rises, the use of waste rubber in civil engineering is gaining attention. This study proposes a method to modify waste rubber using polyvinyl alcohol (PVA) to enhance its material properties and expand its applications. A dataset was created focusing on the mechanical strength of cementitious materials incorporating PVA-modified waste rubber, and multiple machine learning methods were used to develop regression prediction models, particularly evaluating the support vector regression (SVR) model. Results show that the SVR model outperforms others, achieving mean squared errors of 1.21 and 0.33, and mean absolute errors of 2.06 and 0.15. Analysis indicates a negative correlation between waste rubber content and the water-to-cohesive ratio (w/c) with strength indexes, while a positive correlation exists between curing age and PVA. Notably, waste rubber content significantly affects strength. The mechanical strength of cementitious materials was notably enhanced by PVA-modified waste rubber, likely due to PVA's dispersion and bridging effects. This study presents a novel approach to sustainably recycle waste rubber, highlighting its potential in construction materials.