Title:
Relative Compressive Strength of Concretes under Elevated Temperatures
Author(s):
Erdogan Ozbay and Mohamed Lachemi
Publication:
Materials Journal
Volume:
109
Issue:
2
Appears on pages(s):
165-175
Keywords:
compressive strength; elevated temperature; high temperature; modeling; neural network.
DOI:
10.14359/51683703
Date:
3/12/2012
Abstract:
The effects of aggregate type (AT), test type (TT), maximum exposure temperature (MET), and compressive strength before fire (CSBF) on the relative strength (RS) of concrete have been thoroughly explored. For this purpose, the worldwide database of experimental research results of temperature-dependent properties of unreinforced concrete compiled by Knaack et al. was used to generate training and testing data sets to construct closed-form solutions for the neural network (NN). The AT, TT, MET, and CSBF were used as input parameters and relative compressive strength was used as the output parameter of the NN model. The results were used to develop predictive relationships for compressive strength loss under fire and for comparison with the models proposed by ACI 216.1-07 and Knaack et al. A sensitivity study of the parameters was also carried out, and the underlying complex nonlinear relationships among these parameters were investigated. Through the results of modeling and simulation studies, it was demonstrated that a single NN-based explicit equation can be obtained and used to define the RS of concrete exposed to fire.