A Machine Learning Approach to Prediction of the Compressive Strength of Segregated Lightweight Aggregate Concretes Using Ultrasonic Pulse Velocity

Lightweight aggregate concrete (LWAC) is an increasingly important material for modern construction. However, although it has several advantages compared with conventional concrete, susceptible to segregation due the low density of incorporated aggregate. The phenomenon can adversely affect mechanical properties LWAC, reducing its compressive strength and durability. In this work, machine learning techniques are used study influence LWAC on strength, including K-nearest neighbours (KNN) algorithm, regression tree-based algorithms such as random forest (RF) gradient boosting regressors (GBRs), artificial neural networks (ANNs) support vector (SVR). addition, a weighted average ensemble (WAE) method proposed that combines RF, SVR extreme GBR (or XGBoost). A dataset was recently predicting employed in experimental study. Two different types lightweight (LWA), expanded clay coarse natural fine limestone aggregate, were mixed produce LWAC. To quantify ultrasonic pulse velocity adopted. Numerical experiments carried out analyse behaviour obtained models, performance improvement shown models reported previous works. best GBR, XGBoost method. good choice weights WAE allowed our approach outperform all other models.