Perspectives of computer-aided and AI-augmented first-principles-based modeling

Abstract

This perspective paper highlights some of the features of the classes of mathematical models commonly known as data-driven models, first-principles models, and hybrid models. While mathematical models are widely used in various activities in chemical and related engineering domains, in the era of computers and artificial intelligence, their importance and role have assumed increased significance. The demand for discovering novel and better products, made through sustainable manufacturing, demands significant reduction of resources used and time to market. This and other related challenges could be met through versatile models, efficient modeling approaches, and reliable and rapid problem solution through computer-aided model-based systems. The paper highlights some of the related issues through a focused review of selected topics. It explores how artificial intelligence techniques, including machine learning, physics-informed neural networks, and agentic artificial intelligence can accelerate critical modeling tasks such as parameter estimation and solving of inverse problems. Specifically, the paper highlights how process systems engineering-based modeling frameworks augmented with artificial intelligence tools can significantly reduce the time and human resources required to transition from a physical concept to a solved model. Rather than pursuing fully automated modeling, we advocate a synergistic “human-in-the-loop” paradigm where, for example, artificial intelligence would accelerate repetitive computational, retrieval, and coding tasks, while human intelligence would provide critical system descriptions, boundary definitions, and validation. Finally, the paper evaluates these collaborative roles across key modeling workflows and offers perspectives on navigating core challenges such as structural consistency, hybrid model integration, and uncertainty management.