Artificial neural networks as a surrogate model for simulating biomass enzymatic hydrolysis step

• Grant number: 25/01916-6
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Start date: May 31, 2025
• End date: August 30, 2025
• Field of knowledge: Engineering - Chemical Engineering - Chemical Process Industries
• Principal Investigator: Antonio José Gonçalves da Cruz
• Grantee: Augusto Demetrio Canutilho
• Supervisor: Carina Loureiro da Costa Lira Gargalo
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Institution abroad: Technical University of Denmark (DTU), Denmark
• Associated to the scholarship: 24/13885-5 - Optimization of second-generation ethanol production employing machine learning techniques, BP.IC

Abstract

Currently, 80% of the global energy consumption comes from fossil fuels. In this context, renewable fuels, or biofuels, provide an environmentally friendly alternative, as they are derived from biomass and can help reduce carbon dioxide emissions. Lignocellulosic feedstocks, which are rich in carbohydrates and relatively low-cost, have emerged as a promising source for biofuel production. However, the monomeric sugars in these materials are not readily available due to their presence in a complex structure composed of cellulose, hemicellulose, and lignin. Glucose and xylose, the monomeric sugars that are converted into ethanol by yeasts (known as second generation ethanol), are the primary units of the cellulose and hemicellulose, respectively. To extract these sugars, it is necessary to disrupt the lignocellulosic complex through two key operations: pretreatment and enzymatic hydrolysis. This project aims to develop a surrogate model based on artificial neural networks to model and simulate the enzymatic hydrolysis process. Literature data on lignocellulosic pretreatment and hydrolysis will be used to identify the main variables influencing the process, enabling the creation of a representative model of the process. The mathematical model will be implemented in Python software, with a graphical interface developed to facilitate the user interaction, visualization, and also continuous model improvement.