Prediction of Carbon Steel Corrosion Rate Based on an Alternating Conditional Expectation (Ace) Algorithm

Abstract

Based on dynamic corrosion experiments, we propose a new model for predicting corrosion rate that is based on an alternating conditional expectation (ACE) algorithm. This model lets us more accurately predict the corrosion rate for a broad range of temperatures, pH, and concentrations of Ca²⁺, HCO₃⁻, Mg²⁺, Cl^–, SO₄^2 − ions. Based on tests performed on a testing sample group, we have confirmed the reliability of the model and have also demonstrated its high accuracy. Sensitivity analysis based on a rank correlation coefficient revealed that the major factor influencing the corrosion rate of N80 steel is the pH value. We have also carried out a comparison analysis of the results obtained when using the ACE algorithm and the results obtained when using a backpropagation neural network (BPNN) and the support vector regression (SVR) method. As a result, we found that the model based on the ACE algorithm is more accurate than other currently used models.