Prediction of stress intensity factor for cracked steel W-beams based on artificial neural network (ANN)

Abstract

In this study, the stress intensity factor K for web cracks of a Wshaped steel beam was predicted using an artificial neural network (ANN). The input parameters of the ANN model in this study were as follows: the area ratio between the beam flange and the web, the crack length, and the crack eccentricity. In the present analysis, a very high correlation between the K values predicted by the ANN and the FE database was observed, as the correlation coefficient R of the overall model was greater than 0.99. Furthermore, the obtained ANN model was also independently validated with the results from a steel beam outside the FE database to evaluate the generalization ability of the ANN method in the study.

Keywords: W-shaped beam; finite elment method; artificial neural network (ANN).