Predictive Analytics in Construction: Multi-Output Machine Learning Models for Abrasion Resistance

This study aims to accurately predict abrasion resistance, measured through the Los Angeles (LA) abrasion test, and modulus of elasticity, assessed using the Micro-Deval Abrasion (MDA) test, to support structural integrity and efficient material use in construction projects. We applied multi-output machine learning models—specifically Linear Regression (LR), Huber, RANSAC, and Support Vector Regression (SVR)—to predict LA and MDA values based on primary input parameters, including Uniaxial Compression Strength (UCS), Point Load Index (PLI), Schmidt Hammer Rebound (Sh_h), and Ultrasonic Pulse Velocity (UPV). The experimental work involved assessing model performance using metrics such as Mean Absolute Error (MAE), R-squared (R2), and Mean Squared Error (MSE). Linear Regression demonstrated superior predictive accuracy, achieving 94% for R2 with an MAE of 0.21 and MSE of 0.09 for LA predictions and 92% for R2 with an MAE of 0.24 and MSE of 0.11 for MDA predictions. These results underscore the potential of machine learning techniques in accurately predicting critical material properties, offering engineers reliable tools for optimizing material selection and structural design. This research contributes to the advancement of construction practices, promoting the development of durable and efficient infrastructure.

This creative work has been published under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license which allows copying, and redistribution as well as adaptation of the original work provided appropriate credit is given to the original author and the conditions of the license are met.