Coupled propagation behavior of multiple fatigue cracks in welded joints of steel-bridge.

Fatigue cracking in welded joints is a critical problem in bridge engineering, especially for aging long-span steel bridges. The crack size and density are increasing gradually with the bridge aging process, where the crack interaction will accelerate the crack propagation or even short cracks merging into a long crack. However, the mechanism underlying the coupled propagation behavior of multiple cracks remains unclear. In this study, the coupling effect of multiple cracks was investigated based on numerical simulations. Subsequently, the numerical result was verified by a full-scale segmental fatigue experiment. In order to overcome the time-consuming iterative computations, the back propagation (BP) neural network was utilized to predict the effective coupling spacing between multiple cracks. Finally, a new crack reconstruction method was proposed to simplify crack merging processes. Results indicate that the coupled propagation behavior is significantly impacted by the effective crack spacing. The coupling effect is more significant at the near-end point compared to other feature points. In addition, the cracks with similar sizes have greater coupled effect, which is the worst-case scenario for the multiple fatigue crack problem. The experimental study demonstrates that the growth rate of the near-end point is 1.53 times that of the far-end point and pre-made cracks merge together successfully. By utilizing the BP neural network, the relative error of predicted effective spacing of two cracks is 3.52%. Compared to existing design specifications, the new reconstruction method provides a more reliable result for the fatigue life prediction of the welding seam with multiple fatigue cracks. • Sensitive parameters for multiple crack coupling effects were investigated. • The coupling effect has a significant effect on the near-end point of the crack. • The coupling effect and effective spacing increase with crack shape ratio and size. • A crack reconstruction method is proposed based on the BP neural network. [ABSTRACT FROM AUTHOR]