Abstract: The fluorination furnace is a critical piece of equipment within nuclear material production lines. Its process optimization design involves intricate physicochemical processes and variable equipment performance. To achieve real-time, rapid computation for the digital twin multiphysics coupling model of the fluorination furnace, a digital twin lightweight surrogate modeling method was proposed based on generative adversarial networks (GAN) for multiphysics-coupled calculations of the furnace in this paper. Building upon the physical and chemical reaction mechanisms governing the fluorination process within the furnace, a comprehensive COMSOL multiphysics model was first constructed. This model facilitates virtual simulation of the fluid dynamics, heat transfer, and chemical reaction characteristics under diverse operating conditions. However, the computational efficiency of this detailed multiphysics coupling simulation is inherently low. To address this computational challenge, a novel method for constructing a lightweight surrogate model for the fluorination furnace’s multiphysics digital twin was introduced, leveraging Generative Adversarial Networks. The core objective of this method is to significantly enhance computational efficiency while maintaining high fidelity in reproducing the distribution results of the physical fields (such as temperature and gas concentration). This approach dramatically reduces the demands on hardware computational performance and lowers the deployment and application costs associated with utilizing the digital twin model in practical settings. Validation testing demonstrates that the proposed lightweight surrogate model achieves excellent predictive accuracy. The deviation between its computational results and those generated by the original high-fidelity multiphysics coupling model is minimal, as evidenced by root mean square error (RMSE) values: specifically 0.008 for the temperature field and 0.011 for the gas concentration field. These low RMSE values confirm the surrogate model’s capability to reliably approximate the complex multiphysics behavior of the fluorination furnace with significantly reduced computational overhead.