Probabilistic Approach to Predicting Fatigue Lives of Corroded 2024-T3

Abstract

The objective of this work was to develop a probabilistic approach to predict fatigue lives of corroded 2024-T3 aluminum tensile specimens. An experimental program was established to corrode fatigue specimens made of a 2024-T3 sheet separately on the longitudinal-transverse (LT) and longitudinal-short (LS) metallurgical planes in a 3.5% sodium chloride solution for 6 and 8 days. The specimens were fatigue tested to fracture, and the corrosion pits that nucleated fatigue cracks were analyzed with an electron microscope. Corroded material from the fractured specimens was polished on the short-transverse plane and the pits were viewed and photographed under a light microscope. The largest 10% of the collected pits were fit to Gumbel extreme-value distributions. These distributions were used in a Monte Carlo simulation in which 1000 pit areas were selected, modeled as circular surface or corner cracks, and treated as initial flaw sizes. Fatigue life predictions were based on crack propagation life by use of the initial flaw sizes from the Monte Carlo simulation. Predicted cumulative distribution functions of fatigue life were within 22% of the experimental cumulative distribution functions for the LS specimens. The method predicted a reasonable distribution offatigue lives for the 6-day LT specimens but consistently underpredicted the experimental distribution of fatigue lives for the 8-day LT specimens.