Corrosion inhibition of mild steel by some sulfur containing compounds: Artificial neural network modeling

Abstract

Advances in corrosion inhibition studies revealed that the relationship between structure of the inhibitor molecules and their inhibition efficiencies is essentially a non-linear and highly complex process particularly out of reach of classical statistical modeling techniques. The non-linearity of the corrosion process forced us to look for other solutions to track this complex process. Application of Artificial Neural Networks (ANNs) may provide better and more comprehensive results. In this work ANNs were used to predict the inhibition efficiencies of ten sulfur containing compounds on the corrosion of mild steel in hydrochloric acid solutions. A (6-3-1) network was adopted to predict the corrosion inhibition efficiencies of the sulfur containing compounds. The descriptors (inputs) were obtained using quantum chemical calculations. Highest occupied molecular orbitals, EHOMO), lowest unoccupied molecular orbitals, (ELUMO, energy gap, (ELUMO-EHOMO), molecular area, molecular volume and total dipole moments were selected as the ANN inputs to predict the corrosion inhibition efficiencies (output).