Probabilistic S-N curves for CFRP retrofitted steel details

Abstract Experimental fatigue data of non-cracked open-hole steel specimens retrofitted with bonded carbon fiber reinforced polymer (CFRP) are collected from the literature and assessed in this paper. The specimens are representative of steel plates in tensile riveted or bolted connections. The CFRP substantially extends the fatigue life of specimens at stress range Δ σ 220 M P a where the fatigue data points converge towards a fatigue limit. These are promising results since most of the ancient metallic riveted assemblies are subjected to stress levels below 150 MPa. The fatigue life extension ratio decreases when the stress level increases due to the fatigue degradation of the bonding joint and/or the CFRP. The fatigue data of specimens retrofitted with CFRP follow a nonlinear trend and code-based linear S-N curves with a slope m = 3 do not capture this behavior. The commonly used methods that assume the fatigue life extension ratio as constant are not adequate to predict the full fatigue life of CFRP retrofitted specimens. To assess the fatigue behavior of the CFRP retrofitted specimens, probabilistic methods are adopted in this study using: the Normal distribution; the two-parameter Weibull; the Gumbel; and the three-parameter Weibull model developed by Castillo & Fernandez-Canteli (CFC). Regardless of the assumed distribution function, linear probabilistic S-N curves (P-S-N) are not adequate to model the complex behavior of retrofitted specimens with CFRP. The nonlinear CFC model fitted fairly well the different trends of the fatigue data including the fatigue life at high cycle regime and captured the transition between medium–high cycle range. P-S-N curves derived using the CFC model are proposed for CFRP retrofitted specimens.