Work

Machine learning-based analysis of solvothermal lignin and lignocellulose liquefaction

Year: 2024
Type: dissertation
Abstract: Lignin is an abundant bio-polymer found in all plant matter in significant quantities (8~33% by weight), and its chemical structure is comprised interlinked aromatic units via ether and carbon-carbon chemical bonds.Because of its renewable nature it has been suggested as a possible source of renewable aromatic chemicals that could significantly contribute to the decarbonization of liquid fuel and aromatic chemical production.Among the existing depolymerization methods, solvolysis is seen as one of the most viable candidates due to its ease of application and easy scalability.However, in spite of the large amount of works found in the literature, how these solvolysis processes work remains unclear, with multiple authors claiming that the method they present is the "best" choice and put forward suggestions on what future research should be focused on.This dissertation presents a machine learning-based approach to study and try to understand how different experimental factors influence the outcome of lignin solvolysis experiments, by using data captured from literature and training models that predict experimental performance metrics such as the yield of liquid products and solid residues obtained from solvolysis experiments, focusing first in heterogeneously catalyzed, then homogeneously catalyzed solvolysis and finally studying the possibility of using the measured higher-heating-value (HHV) from bio-oil as an alternative experimental performance metric.The results indicate that catalysts properties (both for heterogeneous and homogeneous) play a comparatively small role in predicting the outcome of solvolysis experiments, and that temperature and reaction time is overwhelmingly more important.The yield of liquid products (bio-oil) was deemed to be an unreliable experiment performance metric at times due to it being influenced by the choice of chemical work-up steps used in a given study, further complicating comparison across studies. Chapter 4: Prediction of higher heating values in bio-oil from solvothermal biomass conversion and bio-oil upgrading given discontinuous experimental conditions
Authors: CASTRO, GARCIA Abraham, Castro Garcia, Abraham
Language: English
Open Access status: green
OpenAlex ID: W7030107148
Cites: 162
Cited by:
Related to: 10