Heuristic machine cognition to predict fire-induced spalling and fire resistance of concrete structures

Abstract

Abstract The exceptional behavior of concrete under fire conditions is often jeopardized by concrete's propensity to spall. While published works seem to agree on the complexity and randomness of fire-induced concrete spalling, attempts carried out in the past few years continue to be short of developing a systematic methodology that enables accurate prediction of this phenomenon. Unlike previous works, this study aims at understanding fire-induced spalling of concrete through a modern perspective. In this study, Machine Cognition (MC), a branch of Machine Intelligence (MI), is used to derive expressions able of accurately tracing fire response of concrete structures. These expressions take into account geometric, material, and specific features/properties of reinforced concrete (RC) columns in order to predict occurrence and intensity of fire-induced spalling as well as to evaluate fire resistance of such structural members. The derived expressions implicitly account for high-temperature properties of concrete and steel, and thus do not require input of such properties nor special simulation environment. These expressions, arrived at through observations obtained from actual fire tests, have been calibrated and validated for fire exposures far exceeding that of 4 h and their prediction capability was examined against commonly used calculation methods such as those adopted in European and Australian codes.