Unlocking the potential of FTIR for corrosion inhibition prediction exploiting principal component analysis: Machine learning for QSPR modeling

Abstract

Background

Aiming to find a quantitative structure-property relationship (QSPR), researchers have implemented a variety of approaches to gain prediction power for the corrosion inhibition of chemicals. This work aims to develop a practical approach for the connection of features extracted from FTIR exploiting principal component analysis to corrosion inhibition.

Methods

To this end, an artificial neural network (ANN) was utilized to build a machine learning model based on electrochemical impedance spectroscopy (EIS) data of ten known organic synthetic substances in three different acid solutions on mild steel. Inhibition power and all EIS data points were separately taken into consideration as objectives of the model. Then, the performance of the ANN was validated by employing an unseen plant extract during the training stage.

Significant findings

The predicted data was profoundly conformed with the measured data for the plant extract, especially by employing EIS data points as objectives of the model with R²: 0.95123 and 0.94721, respectively, for the impedance and phase angle Bode plots, clarifying the robustness of the built QSPR model to predict unseen data.