Innovative applications and technological breakthroughs of artificial intelligence in complex multiphase flow systems: A review

This review systematically evaluates the transformative impact of artificial intelligence (AI) on the modeling, analysis, and control of complex multiphase flow systems. Multiphase flow has long been regarded as a significant challenge in the field of fluid mechanics due to its strong nonlinearity, multiscale coupling, and dynamic interface behavior. Traditional methods, including empirical correlations, theoretical modeling, and high-fidelity numerical simulations, frequently face limitations in adaptability, predictive accuracy, and real-time performance under complex, heterogeneous, and dynamic conditions. Recent advances in machine learning, deep learning, and physics-informed modeling have reshaped this landscape, enabling data–physics hybrid models, accelerated simulations, flow regime classification, intelligent parameter prediction, and closed-loop process control. This paper reviews recent progress in three key areas. First, it explores AI-enhanced modeling and simulation techniques, with a focus on data-driven models, physics-informed neural networks (PINNs), and hybrid modeling strategies that integrate prior knowledge. Second, it highlights AI applications in flow field analysis, including structural pattern extraction, intelligent forecasting of key variables, and reinforcement learning-based dynamic regulation. Third, it examines representative industrial applications across the energy, petrochemical, environmental, and bioprocessing sectors and presents a comparative analysis of leading AI methods (e.g., CNN, GNN, GAN, PINN, and RL) in terms of their functional capabilities and engineering suitability. Despite these advancements, the practical deployment of AI in multiphase flow systems still encounters major obstacles, such as the scarcity of high-quality labeled data, limited interpretability and physical consistency of models, constraints in edge computing environments, and insufficient generalization across operating conditions. To address these limitations, the paper outlines future research priorities, including multimodal data fusion, mechanism-informed learning frameworks, integration with digital twins, lightweight edge AI algorithms, and the development of open-source, cross-disciplinary collaboration platforms. AI is increasingly driving the shift of multiphase systems toward intelligent, adaptive, and fine-grained management, offering new momentum for achieving critical goals in energy security, carbon neutrality, and smart manufacturing. The intelligent evolution of multiphase flow science is expected to play a central role in the forthcoming transformation of scientific computing and industrial systems.