In this work, the localized corrosion of AA7075-T651 as a function of environment is empirically modeled using neural network approaches. The modeling approach is divided into three stages: pit initiation, corrosion mode differentiation, and propagation. This study characterized the effect of temperature (0-60°C), pH (2.5-12.5), [Cl-] (0.01-0.6M NaCl), electrochemical potential (-780 to -640mVSCE), orientation (LS, ST), time (1-720h), and alloy microstructure on each stage. Examination of metastable pits during potentiostatic testing was employed to study pit initiation, while differentiation and propagation were investigated by extensive observation of corrosion after immersion exposure using optical profilometry and scanning electron microscopy (SEM).
Pit initiation rate, λ, was shown to have an exponential dependence on electrochemical potential, E, a logarithmic dependence on [Cl-], an exponential dependence on temperature, and peaked at intermediate pH. Additionally, λ scaled with the number density of intermetallic particles. Pitting event intensity (related to the peak current sustained) was also an important parameter in this study. Combining the effects of pit initiation rate and event intensity, [Cl-] and temperature had a profound impact on the total damage incurred (the total charge passed from pitting events). It was established that pit location was determined by the alloy microstructure, while the environment determined the severity of damage. Unlike previous studies that indicate λ decreases exponentially with time, the pit initiation rate was shown to remain constant with time for most of the conditions studied. A trained neural network model was able to accurately predict λ as a function of environmental variables. The neural network was able to reflect previously observed trends in this work and in the literature.
Pitting was determined to be the main mode of localized attack in this study, since no intergranular corrosion (IGC) was detected. Grain boundary attack (distinguished from IGC by presenting on the exposed face rather than in the cross-section), uniform corrosion, and pits forming at both anodic and cathodic particles were observed. Circumferential pits were shown to occur when the reduction current on cathodic particles was ~10 times larger than icorr in a given environment. Solution pH was shown to have a large effect on the corrosion morphology, while temperature and exposure time affected the severity of damage. [Cl-] and orientation had a limited effect on the damage accumulation in this alloy. The number of pit sites after 720h exposure was determined to be ~100-200/mm2 in all environments investigated.
Tested neural network models were able to predict, not only the maximum pit depth and diameter as a function of environment, but also entire pit depth and diameter distributions. Pit growth kinetics varied depending on the exposure conditions, but most environments followed t1/3 kinetics, which is within the range reported in the literature.