On November 26, Professor Luo Kaihong, a Fellow of the Royal Academy of Engineering, attended our university's "Frontiers in Intelligent Construction and Smart City Research" forum, delivering an insightful lecture titled "Cross-Scale Modeling and Simulation Based on Physics and Data-Driven Approaches" to faculty and students. The lecture explored the deep integration of computational fluid dynamics (CFD), molecular dynamics, the Boltzmann method, and artificial intelligence in scientific research, and jointly discussed the future trajectory of China's scientific research capabilities and the Fourth Industrial Revolution.
Professor Luo Kaihong emphasized that modern science, since its foundational work by Newton, has consistently prioritized the discovery and application of physical laws. Computational fluid dynamics (CFD) serves as an indispensable research tool in fields like aerospace, energy systems, and marine engineering by solving equations derived from these laws. With the integration of AI algorithms, high-density computational scenarios such as combustion and turbulence have achieved orders-of-magnitude acceleration. However, issues like error accumulation and long-term stability remain critical challenges requiring attention.
In molecular-level simulations, Professor Luo Kaihong highlighted that classical molecular dynamics (MD) and reaction molecular dynamics (RMD) focus on "atomic/molecular motion descriptions" and "chemical reactions" respectively, accurately reproducing fluid transport, material deformation, interfacial heat transfer, and even electrochemical reactions. Quantum mechanics, as a first-principles approach, provides deeper insights but incurs high computational costs. The lattice Boltzmann method excels in modeling collisions and migration between molecular clusters, demonstrating remarkable performance in simulating complex mesoscopic phenomena such as gas-liquid-solid interfaces, boiling heat transfer, and battery charging/discharging, thereby offering novel predictive methods and innovative approaches for new energy equipment design.
Regarding whether AI+physics bridges gaps or amplifies errors, Professor Luo Kaihong's perspective is that machine learning has demonstrated remarkable capabilities in tasks like fluid simulation, chemical computation acceleration, and parameter inversion, effectively compensating for the limitations of traditional numerical methods in sparse parameter spaces. However, the presentation cautioned that AI agent models carry the risk of "uncertainty drift," where prolonged simulations may amplify systematic errors. To achieve "trustworthy data-driven simulations," these models must be deeply integrated with physical constraints and conservation laws.
Reflecting on the Fourth Industrial Revolution, Professor Luo Kaihong noted that while the first three revolutions were driven by steam engines, electricity, and information technology respectively, the fourth is likely propelled by a dual engine of "AI + new energy". Artificial intelligence boosts R&D efficiency, while new energy reshapes industrial foundations and supports sustainable development. Seizing this opportunity requires a tripartite synergy breakthrough in algorithms, computing power, and industries.
Professor Luo Kaihong highly praised the strength of the motherland and presented existing data to prove that China ranks first globally in indicators such as electricity generation, vehicle and ship production and exports, and robot usage; the annual publication of high-level papers and the total number of scientific research talents both lead the world, and the innovation index has risen to the top ten globally. However, the scientific research output is "high in points but low in scope," and the proportion of innovative enterprises is insufficient, which constrains the "last mile" from technology to industry. He called for accelerating the construction of a complete ecosystem of "basic research—technology development—scenario implementation—large-scale manufacturing," balancing production and consumption, domestic demand and exports, and striving to achieve a decisive victory in the new round of industrial revolution.
During the final exchange session of the conference, Professor Luo Kaihong discussed with students that AI does not replace physical models but amplifies their boundaries; traditional simulation and modeling need to rely on AI to break through the "ceiling." In the next decade, the industry urgently needs to develop an "AI framework embedded in physics," achieving "trustworthy simulation and modeling" across scales and disciplines while ensuring conservation, symmetry, and stability. If China can take the lead in completing the full-chain layout of "algorithm-software-equipment-industry," it is expected to occupy the commanding heights in the new round of industrial revolution.
