Mediterranean Congress on Chemical Engineering (MeCCE)

In the early twenties, sustainability moved to the center of global agendas, with governments and companies committing to cut emissions, use resources more efficiently and mitigate climate, biodiversity and human-rights risks. Financial institutions pushed for evidence of action, and new regulations accelerated disclosure.

Today, geopolitical tensions, economic uncertainty, energy constraints and the rise of AI have shifted attention, while changes in reporting rules have slowed momentum. Yet this moment may be critical: this lecture will explore whether acting now through circularity, resource efficiency and smarter design can drive innovation, competitiveness and solutions that meet planetary needs.

Due to the Industry 4.0 revolution, several process challenges must be tackled: (i) How to quickly develop new models, (ii) how to automatically update them, (iii) how to estimate uncertainties

To address them, IFPEN has developed two strategies based on AI. They depend on the extent of knowledge about the chemical process. If a model structure is known, transfer learning is used. Otherwise, hybrid model strategies are proposed.

IFPEN is conducting several research projects on knowledge transfer methods, particularly Bayesian approaches.

IFPEN has developed a physics-informed hybrid modeling framework that integrates mechanistic knowledge with a neural network through automatic differentiation, differential residual loss minimization, and physical boundary constraints.