Explainable machine learning using real, synthetic and augmented fire tests to predict fire resistance and spalling of RC columns

This paper presents the development of systematic machine learning (ML) approach to enable explainable and rapid assessment fire resistance fire-induced spalling reinforced concrete (RC) columns. The developed comprises an ensemble three novel ML algorithms namely; random forest (RF), extreme gradient boosted trees (ExGBT), deep (DL). These are trained account for a wide collection geometric characteristics material properties, as well loading conditions examine performance normal high strength RC columns by analyzing comprehensive database tests comprising over 494 observations. is also capable presenting quantifiable insights predictions; thus, breaking free from notion 'blackbox' establishing solid step towards transparent ML. Most importantly, this work tackles scarcity available proposing new techniques leverage use real, synthetic augmented test has been calibrated validated standard design exposures one, two, four-sided thus; covering range practical scenarios present during incidents. When fully deployed, can analyze 5,000 in under 60 seconds providing attractive solution researchers practitioners. presented be easily extended evaluating other structural members varying hence paves way modernize state research area practice.