Consistent Estimate of Brace Fatigue Life Using Axial Deformations Evaluated from Nodal Displacements

Abstract

Characterizing the behavior and assessing the performance of concentrically braced frames depends on adequately simulating force redistributions following brace fracture. Commonly, brace models with beam-column elements estimate fatigue life using the Coffin-Manson relationship and Miner’s rule. These “fatigue-material” models accumulate damage based on strains evaluated at each fiber. However, strains evaluated at the fiber level depend on the choice of integration points and subelements used to discretize the brace, leading to variations in estimates of brace fatigue life across studies. Moreover, dependence on the mesh resolution means that the fatigue-material parameters need to be carefully calibrated to the experimental data based on the chosen brace discretization. To achieve consistent strains at the fiber level, some have suggested using 10–20 force-based beam-column subelements. However, force-based elements were formulated to explicitly satisfy the equilibrium between the element and section forces without the need for an overly fine mesh. A “fatigue-member” model is proposed based on axial deformations evaluated at the nodes normalized by the undeformed effective length, which needs only two force-based subelements to consistently estimate brace fatigue life based on the Coffin-Manson relationship and Miner’s rule. The fatigue-member uses deformations calculated from the nodal displacements rather than fiber-level strains, resulting in a low-cycle fatigue model that depends less on the model discretization, such as the number and location of integration points or subelements. Since the node-level deformations are more consistent across a choice of mesh resolution, the proposed node-level approach reduces variations in estimates of brace fatigue life across studies, where independent researchers may have preferences or needs for certain modeling decisions. Moreover, the fatigue-member parameters can be derived directly from experimental data since the proposed fatigue model is independent of the mesh resolution.