Predictive analysis of recycled concrete properties at elevated temperatures using M5 pruned rule classifiers

Abstract

The present paper aims to determine the effect of elevated temperature on properties of recycled concrete analytically using the model extraction rule-based M5 algorithms. This approach helps predict both the destructive and non-destructive properties of various concrete mixtures. The dataset employed in the construction of predictive models comprises test data obtained from 35 distinct concrete mix designs. These designs were developed through experimental work, which involved by substituting coarse aggregate with recycled concrete and fine aggregate with copper slag. Weka software, a commonly used tool for machine learning algorithms, is employed for creating these models. Input data corresponding to the concrete mixture’s variables are utilized to predict the model. Results from the model revealed that the predicted data align closely with the experimental data, and correlations between different output parameters can be established. The coefficient of determination, which exceeds 0.8, indicates a strong correlation between various datasets. Overall, the study’s findings demonstrated that M5 rule-based models can generate highly accurate forecasts for the specified mechanical parameters and its performance is evaluated using Taylor diagram.