Prediction model for compressive strength of rice husk ash blended sandcrete blocks using a machine learning models

Abstract

Sandcrete blocks are popular for construction but their production relies on cement, a major greenhouse gas contributor. Rice husk ash (RHA), a waste product, can partially replace cement in sandcrete blocks. This study uses machine learning (ML) to predict the compressive strength of these blocks, in uenced by factors like the ratio of ne aggregate to binder, RHA to binder ratio, water-to-binder ratio, and curing time. The data was collected from published literature on factors affecting compressive strength from various sources and analyzed 795 observations. The analysis showed that strength increases with longer curing but decreases with higher ratios of aggregate-to-binder, RHA-to-binder, and water-to-binder. The data was divided for training and testing ML models. Five algorithms were investigated, and the eXtreme Gradient Boosting (XGB) model emerged as the best for predicting compressive strength. The XGB model strongly correlated with predicted and measured strength, with an R² value of 0.94 for training data and 0.89 for testing. It also displayed lower error metrics compared to other models. XGB's success is due to its ability to handle complex relationships and prevent over tting. This study highlights the potential of ML for predicting the strength of RHA-blended sandcrete blocks.