A formulation and comparison of different shell FE modeling techniques for fatigue life simulation of welded joints

A FE shell model capable of representing a welded structure without any significantly error on its stiffness could be widely applied to dynamic problems in which the structural stress method (hot spot approach) is employed for fatigue analysis. The scope of the present work is to formulate a modeling technique capable of doing so. In order to accomplish it, a parametric optimization for simulating welded structures using shell elements is made, the design variables in the proposed formulation are defined as the weld leg length and thickness of the shell element representing the weld fillet. The main goal of the optimization was to find a range of thickness/leg length which would not change significantly the first natural frequencies, and still deliver results similar to the ones obtained by a solid model. Sequential linear programming optimizations are performed in a T-shaped structure, with constant section and different plate thicknesses and depths. Once the optimal parameters are found, two different modeling techniques are presented and compared with three well established methodologies presented in standards and the literature. The differences in the results are compared for first natural frequencies, total mass, hot spot stress and fatigue life.