ANN based fire resistance prediction of FRP-strengthened RC slabs with fireproof panel including air layer

Abstract

In this study, an Artificial Neural Network (ANN) model to predict the fire resistance of Fiber Reinforced Plastic (FRP) strengthened Reinforced Concrete (RC) slabs with fireproof panel including air layer exposed to fire is developed. FRP-strengthened RC slabs with fireproof panel including air layer has a structure that improves structural performance and fire resistance by placing FRP and a fireproof panel including air at the bottom of the RC slab. First, fire test is performed using the air layer as a variable, and the validity of the software ABAQUS heat transfer analysis is verified by comparing it with the fire test results. Through this heat transfer analysis, the temperature distribution of the slab is determined, and 1521 data sets are obtained by calculating the strength reduction coefficient and elastic coefficient reduction coefficient of each material. The data consist of eight input parameters: the type of fire curve, fire exposure time, insulation thickness, air layer thickness, concrete covering thickness, and FRP thickness. The output parameter is the value of the strength and elastic modulus reduction coefficient for each material. The average values of MSE, RMSE, R, and SSE for all output parameters, which are performance indicators of the ANN model, are 0.00202, 0.03978, 0.98235, and 1.10445, respectively, and the developed ANN model has high computational accuracy and high generalization ability. Therefore, the ANN model developed here can determine the mechanical properties of each material in FRP-strengthened RC slabs with fireproof panel including air layer exposed to fire and evaluate the fire resistant without complex calculations. Through contribution and sensitivity analysis, the effect of input data on output data is confirmed.