Prediction of Explosive Spalling of Heated Steel Fiber Reinforced Concrete using Artificial Neural Networks

Abstract

Explosive spalling is a severe threat to concrete at high temperature. The addition of steel fibers is believed to be useful to mitigate explosive spalling of concrete. But predicting explosive spalling of steel fiber reinforced concrete remains to be a challenging topic. This paper adopted a popular machine learning approach, i.e., artificial neural network (ANN), to predict explosive spalling of steel fiber reinforced concrete and furthermore study the effect of steel fibers on explosive spalling resistance of concrete. Two ANN models were developed, with ANN1 concrete mix-based and ANN2 concrete strength-based. Twenty groups of heating tests were conducted to validate the proposed ANN models. Both ANN models showed the prediction accuracy of 100%, which demonstrates that ANN is a powerful tool for assessing explosive spalling risk of steel fiber reinforced concrete. A parametric study was also conducted to investigate the effect of steel fibers on explosive spalling resistance of concrete using the well-validated ANN1.