Prediction of Pitting Corrosion Characteristics using Artificial Neural Networks

Abstract

Precorroded steel A-106 B specimens were prepared at different surface roughness. These specimens were immersed in corrosive ferric chloride solution in different concentrations (1.5, 3.0, 4.5, 6.0% wt.) at specified durations to initiate primarily the pitting corrosion. The corrosion pits distribution depend on the corrosive concentration, degree of surface roughness, and immersion duration. The pits were characterized using metallurgical microscope. Also, The pitting characteristics aimed to be predicted by “Artificial Neural Networks” (ANNs). The results obtained of pit quantification by ANNs predictions are shown to be agreed well against experimental values. i.e. R2=0.9839