Journal of Constructional Steel Research 62 (2006) 950–961 www.elsevier.com/locate/jcsr Prediction of rotation capacity of wide ﬂange beams using neural networks Ibrahim H. Guzelbey a , Abdulkadir Cevik b,∗ , Mehmet Tolga G¨ og¨ us ¸ b a Department of Mechanical Engineering, University of Gaziantep, Turkey b Department of Civil Engineering, University of Gaziantep, Turkey Received 19 September 2005; accepted 3 January 2006 Abstract This study proposes Neural Networks (NN) as a new approach for the estimation and explicit formulation of available rotation capacity of wide ﬂange beams. Rotation capacity is an important phenomenon which determines the plastic behaviour of steel structures. Thus the database for the NN training is directly based on extensive experimental results from literature. The results of the NN approach are compared with numerical results obtained by a specialized computer. Available rotation capacity is also introduced in a closed form solution based on the proposed NN model. The proposed NN method is seen to be more accurate than numerical results, practical and fast compared to FE models. c  2006 Elsevier Ltd. All rights reserved. Keywords: Rotation capacity; Plastic behaviour; Neural networks; Explicit solution 1. Introduction The behaviour of a wide ﬂange beam can be generalized into elastic, inelastic and plastic categories as shown in Fig. 1. In any case the failure of the beam will be due to one of the following: local plate buckling of the compression ﬂange, local plate buckling of the web in ﬂexural compression, or lateral–torsional buckling. The plastic behaviour category is of special concern in this study as it permits moment redistribution in indeterminate structures [1]. Plastic analysis and design enables the full cross sectional capacity of a beam to be used by notionally allowing a plastic hinge to form. This hinging occurs when the plastic moment strength, M p , is reached at a discrete point along the beam (i.e. the entire cross section has yielded). At such a location, the cross section can no longer resist increasing moment and hence large rotations occur, with constant resistance, M p , being maintained. In the case of an indeterminate structure, such a scenario allows for moment re-distribution to occur. However, it is critical that in addition to the cross section reaching its plastic moment capacity, the beam must also be ductile enough to maintain M p while continuing to deform (rotate) through a ∗ Corresponding author. Tel.: +90 342 3601200x2409; fax: +90 342 3601107. E-mail address: akcevik@gantep.edu.tr (A. Cevik). sufﬁcient angle so that moment redistribution can take place. A common structural ductility or deformation capacity measure is termed plastic rotation capacity [2]. The estimation of plastic rotation capacity is of signiﬁcant importance for plastic and seismic analysis and design of steel structures. Similarly the moment redistribution in a steel structure also depends on the rotation capacity of the section. Thus the determination of rotation capacity of steel structures becomes an important task. This study focuses on the prediction of available rotation capacity of wide ﬂange steel beams. Theoretical, empirical and approximate methods have been proposed for the determination of available rotation capacity of wide ﬂange steel beams in literature which have been reported by Gioncu et al. [3,4]. In order to ﬁnd how realistic results are, these studies should be compared with experimental tests. Thus an alternative approach for the prediction of rotation capacity of wide ﬂange steel beams using NNs is presented for the ﬁrst time in the literature. Backpropagation NNs are used for the training of the NN model. The results of the proposed NN model based on experimental studies are compared with numerical results and are seen to be very accurate. Moreover an explicit solution of rotation capacity for wide ﬂange beams in terms of geometric and mechanical parameters will be introduced by using the well trained NN parameters. The proposed NN approach is quite accurate, fast and practical compared to the FE approach. 0143-974X/$ - see front matter c  2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcsr.2006.01.003