Structural Correlation and Predictive Model by Machine Learning Regression for Oxidative Degradation Rate of Amine Solution in Post-Combustion Amine-Based CO2 Capture Processes
Carbon capture has been proposed as a viable method of lowering CO2 emissions. The most widely used technology is the use of amines to capture CO2. However, the strategy commonly encounters operational concerns due to the repeated use of an amine, which is amine degradation. Recognizing the amine’s stability prior to the start-up of the CO2 capture plant allows us to develop an effective degradation prevention strategy required to reduce or potentially eliminate degradation from the amine-based CO2 capture process. Furthermore, the structural correlation will aid in amine selection during the initial stages of building a CO2 capture unit, ensuring that only the least degradable amines are used.
The degradation experiment of 27 amines was performed to investigate the degradation rate of each amine type. The assessment of the relationship of amine structure comprising of amino (in non-cyclic and cyclic amines), alkyl (in non-cyclic and cyclic amines), and hydroxyl groups and their linkages, reactivity and amine degradation rate was carried out. Information from this data was further used to developed the mathematical model using multiple linear regression to predict the amine degradation rate. The developed model was produced with a 22% AAD accuracy.
In this work, the degradation rate predictive model will be using machine learning regression with additional assumptions, other measurement criteria, alternative data cleaning techniques, or even incorporated ensemble learning to increase the model’s reliability over our previous degradation model by lowering overall error for greater accuracy.