Reliability analysis of models for predicting T-beam response at ultimate limit response

The aim of the present paper is to compare the current design codes' predictions for reinforced concrete (RC) T-beams with the alternative physical methods - namely, the compressive force path (CFP) method and artificial neural networks (ANNs). Therefore, two databases, for T-beams without stirrups and with stirrups, are developed using the available experimental studies. The comparative study on prediction (obtained from the American Concrete Institute and Eurocode 2, CFP and ANN models) shows that the predictions of the ANN model provide a closer fit to the experimental results; after ANN the predictions of the CFP method are close to the experimental results when compared with the counterpart physical model. Comparative studies are also conducted on the critical parameters for the behaviour of RC T-beams. Furthermore, a non-linear finite-element tool (i.e. Abaqus) is used to validate the prediction of the ANN and CFP model. The crack pattern from Abaqus exhibited the same mechanics, on which the CFP models are based.