Data mining techniques and ultrasonic pulse velocity tests for the assessment of damage levels in concrete exposed to high temperatures and subject to compression

Abstract

A Data Mining (DM) process was applied, aiming to develop a numerical tool for the assessment of the level of damage (measured from a strain parameter) of concrete columns subject to high temperature exposure (or fire). The database used was obtained from an experimental program carried out with measurements of Ultrasonic Pulse Velocity (UPV) through concrete specimens of various steel reinforcement arrangements after exposure to different levels of high temperature (reference, 250 ºC, 500 ºC and 750 ºC). The UPV tests were performed during the execution of the compression tests. Different DM techniques were tested and a sensitivity analysis was made. The obtained results show that DM models, particularly the k-nearest neighbours model, allow an acceptable prediction of the axial strain of concrete elements exposed to high temperatures using UPV data.