Prediction of Corrosion Inhibition Efficiency Via Multilayer Perceptron

Abstract

Corrosion inhibitors have been utilized for various industries globally. The usage of corrosion inhibitors depends on the condition at which the inhibitor is applied, and it can be difficult to select the proper inhibitor for a multitude of conditions. A laboratory testing of the various inhibitors as well as the multiple conditions can be tedious, costly and time consuming. This study attempts to propose a neural network prediction model of the behavior of corrosion inhibitors under various conditions. An artificial neural network is one of the methods that can be utilized to investigate and propose a prediction model based on various factors involved in corrosion inhibition. In this study, a 3 layers multilayer perceptron (MLP) network was used to predict the behaviors of the corrosion inhibitors. The layers were categorized into input layers, hidden layers and out layers. The model employed the dataset obtained from a previous study on the corrosion inhibition of L-valine for mild steel in 5% HCI. The data was segregated into 60% for training, 20% for testing and 20% for validation. It was found that the result simulates a reliable prediction model. The reliability was determined via the Regression value (R) and Mean Squared Error (MSE). High value of R and low value of MSE indicates a reliable prediction model in this study. In conclusion, this study is aimed to develop a reliable neural network prediction model based on the dataset from previous experimental work and compare the impact of different network architecture on the performance of the model.