Characterization of pit growth in Al alloys by neural networks

Pitting damage accumulation was systematically characterized as a function of environmental conditions and exposure time for aluminium alloy 7075 for the purpose of developing pit depth distributions, otherwise know as damage functions. Aluminium alloy 7075-T6 samples were subject to static immersion in solutions of varying temperature (0, 25, 40, 60°C), pH (2.5, 6, 10, 12.5), and chloride concentration (0.01, 0.1, 0.6M [Cl ^-]) and serially removed after 1 hr, 1 day, and 1 month exposure time. In addition to this, two different orientations (LS, ST) were investigated. Pitting was characterized by interferometric laser optical profilometry to quantify the pit dimensions at each condition. Pit depth distributions for each environment were fit to a two-parameter Weibull distribution function. Artificial neural network (ANN) methods were utilized to model both Weibull parameters as well as the maximum pit depth as a function of temperature, solution pH, [Cl ^-], exposure time, and orientation with good correlation. ANN-predicted Weibull parameters were used to predict pit depth cumulative distribution functions (CDF) for independent data sets. The predicted distributions corresponded well with the actual distributions, and the models developed allowed prediction of pit depth damage functions as function of environment and exposure time.