Quantitative research on technological evolution and collaborative division of labor in China’s fusion engineering based on patent big data

Establishing an efficient innovation system that balances the uncertainty of scientific exploration with the rigidity of engineering manufacturing is a critical challenge for nuclear fusion engineering. China’s rapid progress in fusion energy offers a unique empirical case, yet its innovation network exhibits sparse connectivity features that differ from traditional high-density network theories. Based on a longitudinal global patent dataset, this study employs BERTopic modeling and Social Network Analysis (SNA) to quantitatively investigate the technological evolution trajectory and micro-collaborative mechanisms of China’s fusion engineering. Data analysis indicates a significant structural shift in technological focus, transitioning from early verification of inertial confinement and plasma physics to critical reactor core engineering subsystems, specifically superconducting magnet support structures, tritium breeding blankets, and radiation-resistant first-wall materials. Empirical evidence further suggests that this technological modularization corresponds to a distinct governance mode of “modular division of labor”: research institutes act as system integrators managing complex interfaces, universities focus on computational approaches to non-linear physical problems, and enterprises are embedded into vertical supply chains to address precision manufacturing challenges. This asymmetric institutional arrangement based on specialized division of labor prioritizes the efficiency of vertical engineering integration. This study utilizes patent data to reveal the underlying logic of how latecomer nations coordinate the relationship between scientific principles and engineering industrialization in mission-oriented innovation.