A comprehensive description of the failure of materials under cyclic loads is key for their safe operations. Experimental investigations have routinely represented fatigue failure through the S-N curve and the Paris’ law. While in the first case, the failure of specimens results from the growth of microscopic cracks, the second case pertains to tracking the evolution of an existing macroscopic crack. A connection between these two curves has strategic value and yet is not clearly understood. As a result, insights from experiments of fatigue failure are mostly phenomenological. Moreover, with the sensitivity of fatigue life to mean stress, surface finish, etc., traditional safe-life design routinely involves prolonged experimental campaigns.  

During this presentation, we will discuss how the statistical analysis of the roughness of fracture surfaces helps in simplifying, and ultimately shortening, conventional experimental protocols. Recently, statistical fractography has been successfully employed to measure the fracture toughness Kc of failed parts as well as their in-service loading. It relies on the measurements of characteristic length scales of damage processes (e.g. process zone size) from the statistical analysis of material’s fracture surface. 

When it comes to fatigue failure, we observe that such length scales are not constant. Instead, they increase as loading amplitude ΔK increases – an observation that is at odds with the constant size of the process zone considered in standard models. To investigate further this phenomenon, we study the displacement field in the vicinity of a fatigue crack-tip using digital image correlation (DIC). Our experiments carried on Al 7075 alloy reveal the presence of a non-elastic region near the crack tip emerging from the damage activity ahead of the crack. As the fatigue load amplitudes ΔK varies, the characteristic size of this zone varies too, and is found to be in good agreement with the fractography. The existence of a loading amplitude dependent length scale has  interesting implications. Similar to the short-crack theory, it argues for a connection between the mechanism of fatigue involving the initiation of cracks and for the growth of existing cracks.

From an engineering perspective, these results pave the way for determining specimen exhaustive S-N curve reliably from a single fatigue crack growth experiment of high cycle fatigue.  As a proof of concept, we determine the S-N curve from the fatigue experiment of the aluminum alloy. The implications of our work for a simpler characterization of the fatigue failure properties of components is finally discussed.