ANN models for prediction of residual strength of HSC after exposure to elevated temperature

Although high strength concrete (HSC) is becoming popular in building construction around the globe, its performance under high temperature (or fire) exposure is not precisely known. The existing fire-design provisions were developed mostly from the results of fire tests on normal strength concrete (NSC) and thus their applicability to HSC needs to be evaluated because sufficient HSC data is now available. This paper is aimed at developing artificial neural network (ANN) based predictive relationships between the statistically significant parameters and the residual compressive strength of concrete for its application in structural fire design of HSC. The proposed models are based on a large set of experimental data that was collected through an extensive survey of the available tests on HSC after high-temperature exposure. The data was carefully examined and analyzed to identify the statistically significant/sensitive variables and to establish the influence of these variables on the residual strength of HSC. The database was used to check the validity/applicability of the existing design models of codes, standards, guidelines and several researchers. New ANN based residual strength design models for HSC were also proposed.