Machine Learning Prediction Models for Compressive Strength of Concrete with Industrial Waste Materials at Elevated Temperatures

Abstract

Background: Concretes utilized in very high temperature zones, for example, furnaces in chimneys, refineries, and nuclear facilities for the disposal of waste, are exposed not only to temperature changes but also to variations in humidity, all of which drastically impact their Compressive Strength (CS). It has been shown that when the temperature rises, porosity of the material increases, microcracking develops, and cementitious compounds undergo dehydration, resulting in the loss of strength of the material. Methodology: The article presents a survey of the prediction of compressive strength in waste-incorporated concrete through different Machine Learning (ML) techniques at different temperatures and moisture levels. Gaussian Process Regression (GPR) and Gradient Boosting Regression (GBR) have been chosen as the underlying models, while Artificial Protozoa Optimizer (APO)has been applied to tune the hyperparameters. Through SHAP and LIME, model explainability has been increased in order to determine the impact of input variables on the model's prediction.