Application of Regression Models on the Prediction of Corrosion Degradation of a Crude Oil Distillation Unit

Abstract

The crude distillation unit is the most critical elements in the refining process. Moreover, most of the equipment in the distillation unit are made of general carbon steels. Data analysis models, machine learning techniques can predict corrosion degradation rates. We used Pearson’s correlation coefficient and multiple linear regression, to predict the impact of process parameters. Altogether, we have analysed 84 channels of technological parameters, and 22 different types of crude oils. Among the corrosion agents, the chloride content strongly affected the weight loss of coupons, where the highest coefficient was 0.68. The most influential parameter is found to be the pH value. Thus, an estimation method of the pH value is set up to predict the corrosion degradation rate. The regression correlation for estimating the pH value is 0.53 if the corrosion agents are not used, which can be improved to 0.76 if the corrosion agents are also used in the regression analysis.