Fire degradation is an extremely important risk that threatens timber structures. It is therefore normal that timber design codes include provisions for the design and verification of structures under fire loading. Eurocode 5 is no exception to this, but the simplified methods presented in the code show some inconsistencies, and the advanced method is not practical to use for design purposes. Artificial neural networks (ANNs) have the ability to model complex problems and have been used in a variety of construction engineering problems. They can learn from a subset of data, and then they can be used to predict the results for other input parameters. The purpose of this study is to present the possibility of the use of ANNs for the prediction of temperatures in rectangular timber cross sections, under fire exposure.
In this work, a multilayer feedforward ANN has been trained to predict the temperatures within a timber cross section, using as input the size of the cross section, the timber density, the time of exposure and the coordinates of the point within the cross section.
The results obtained clearly indicate that ANN can be used to predict the temperatures in a timber cross section subjected to fire.
ANNs have not been used for the prediction of temperatures in timber cross sections. The use of ANN makes the temperature prediction under a standard fire loading in a cross section extremely easy to implement in any code. These results can be used to calculate the strength of the elements after fire.
Author wishes to acknowledge FCT/MEC for the financial support (FCT/UID/ECI/04450/2013 and SFRH/BSAB/114546/2016).
Emerald Publishing Limited
Copyright © 2019, Emerald Publishing Limited