Integrated Machine\nLearning/FT-IR Framework for Efficient\nSolvent Composition Analysis in Carbon Capture

Solvent composition analysis is key to determining the\nperformance\nof the carbon capture process. Current efforts focus on the optimization\nof sample preparation, analysis time, and method reliability. This\nwork introduces an innovative approach that combines a machine learning\nalgorithm with Fourier transform infrared (FTIR) spectroscopy for\nthe precise and accurate quantification of carbon dioxide (CO<sub>2</sub>) and monoethanolamine (MEA) in carbon capture solvent with\na limited number of samples in the training phase. The framework consists\nof feature selection that combines statistical and domain-knowledge\nanalysis to choose the appropriate wavenumber that represents both\nMEA and CO<sub>2</sub> accurately. Eleven machine learning models\nare compared in terms of accuracy to predict the MEA and CO<sub>2</sub> concentration in solution given the absorbance from the previously\nselected wavenumber. The result shows that extreme gradient boosting\nand adaptive boosting are the best models for predicting MEA and CO<sub>2</sub> concentration, respectively. This study obtained a similar\naccuracy compared to the literature with a data set reduced by a factor\nof 2.63 on average. The framework showcases the synergistic potential\nof ML associated and FTIR spectroscopy, offering a robust methodology\nfor routine analysis of MEA and CO<sub>2</sub> concentration in the\nsolvent.