Framework for the Prediction and Assessment of Corrosion Damages in Amine Systems Using Plant Data, Process Simulation and Data Analysis

Abstract The prediction and assessment of corrosion rates in amine gas treating units include reviewing current and historical environmental components such as amine type, H2S and CO2 loading, and temperature; to identify, trend, and provide corrective actions for potential problems related to streams quality, contamination, or damage diagnosis. This review is completed with data obtained from traditional off-line monitoring methods, such as mechanical integrity reports and analysis of the process streams, to capture the state of criticality of the system. Recent advances in corrosion modeling of amine systems allow integrating this data with numerical modeling to effectively quantity and predict corrosion rates. Numerical modeling is based on empirical models, which are usually limited within the ranges of data used in their development, unlike first-principles models that can accurately extrapolate beyond this range. Furthermore, empirical models may lead to significant errors when extrapolated outside the range of the training data. Therefore, their accuracy can be substantially improved by adding data generated from first- principles models through a sensitivity analysis of process and corrosion-related variables. This work proposes a framework for predicting and assessing corrosion rates in amine gas treating units, using surrogate models that combine process simulation software and plant data. A first-principles model of a simulated amine plant is employed to predict process-related variables, combined with a mechanistic model used to predict corrosion rates. Once the data is collected, exploratory data analysis is employed to quantify the correlation between process and corrosion variables, dimensionality reduction, outliers’ detection and treatment, and model performance evaluation. This framework also provides guidelines for selecting surrogate models predicting process variables and corrosion rates. These models can eventually be coupled with multi-objective optimization algorithms for control purposes.