MARTINS, FRANCISCO1; VASCONCELOS, GRAÇA2; MIRANDA, TIAGO3

1) Associate Professor, ISISE, Department of Civil Engineering, University of Minho, ffm@civil.uminho.pt

2) Assistant Professor, ISISE, Department of Civil Engineering, University of Minho, graca@civil.uminho.pt

3) Assistant Professor, ISISE, Department of Civil Engineering, University of Minho, tmiranda@civil.uminho.pt

The rehabilitation and repair of existing structures requires inspection. This generally includes in situ non-destructive testing. A very economical test is the non-destructive ultrasonic pulse velocity test (UPV). Little information is available in the literature in relation to the use of this technique for the estimation of the tensile strength of materials. Therefore, this paper aims at using artificial neural networks (ANN) in the prediction of the mechanical behaviour of granites under tensile loading. The parameters under analysis are the tensile strength, displacement at peak stress and critical crack opening. For this, experimental results obtained from the physical and mechanical characterization under tension of distinct types of granites are combined and the performance of the developed models using the UPV index alone and combined with other physical parameters is analysed. The results of the ANN models are also compared with respect to the results of regression models. The criteria used to evaluate the predictive performances of the models were the coefficient of correlation (R) and root mean square error (RMSE).

Keywords: granite, tensile strength, ultrasonic pulse velocity, artificial neural networks