• Adeleke, Oluwatobi
• Akinribide, Ojo J.
• Jen, Tien Chien
• Akinwamide, Samuel Olukayode
• Ukoba, Kingsley
• Olubambi, Peter A.

Abstract

Publisher Copyright: © 2024 The Authors ; Ductile iron (DI) samples were immersed in near-neutral, alkaline sodium hydroxide (NaOH), and sodium chloride (NaCl) environments for 180 days. The influence of microstructure on the corrosion resistance of three DI specimens was investigated. Microstructures, electrochemical measurements, and the characterization of the corroded surfaces were analyzed. The experimental results from this study were used to validate a model generated from hybrid adaptive neuro-fuzzy inferences system-particle swarm optimization (ANFIS-PSO) algorithms. The hybrid ANFIS-PSO modelling technique was improvised for a detailed evaluation of corrosion rate of ductile cast iron materials in different environments. The integrated hybrid ANFIS-PSO model revealed a sharp rise in localized corrosion caused by chloride-induced structural deterioration at the nanoscale for some of the grains. The performance results revealed that the fuzzy c-mean (FCM) clustering outperformed other clustering approach in the neuro-fuzzy model. Accuracy values of 92.9% and 93.7% were recorded for the training phase of ANFIS-FCM and ANFIS-PSO-FCM respectively for corrosion rates. The percentage error of the ANFIS-PSO predictions is significantly lower than the ANFIS-standalone prediction. This shows that the ANFIS-PSO with FCM approach is a better model for predicting corrosion rates. This will contribute to the body of knowledge for ductile iron, corrosion, and corrosion modelling using machine learning. ; Peer reviewed

Topics

• impedance spectroscopy
• surface
• grain
• corrosion
• phase
• Sodium
• iron
• cast iron
• clustering
• machine learning