📝 Publications

🎙 Papers Published in Journals

IJNMHF 2024

Efficient modeling of liquid splashing via graph neural networks with adaptive filter and aggregator fusion

J. Nan, P. Feng, J. Xu, F. Feng

Project | SCI Q1, Scopus CiteScore top 10% International Journal of Numerical Methods for Heat & Fluid Flow, Vol. 34, No. 2, pp. 200-222. Highlights: * Introduced an innovative GNN framework (FEGNS) to model liquid splashing dynamics with high accuracy and efficiency (AI for Science). * Achieved a 30.3% improvement in simulation accuracy and a 51.6% gain in speed compared to traditional CFD methods. * Validated on extensive datasets from the German Institute and TUM, demonstrating robust generalization and superior performance.

• J. Nan, P. Feng, J. Xu, F. Feng. (2024). Advanced Prediction of Microfluidic Flow in Medical Pipelines Using Graph Neural Networks. International Journal of Numerical Methods for Heat & Fluid Flow. (Under Review)
  • Highlights:
    • Proposed a novel GNN approach for predicting microfluidic flow in complex medical pipelines, outperforming traditional CFD methods.
    • Successfully reduced computational time while maintaining high prediction accuracy for both steady-state and transient flow behaviors.

👄 National Software Copyright & Patent

• J. Nan et al. (2024). Neural Network-driven SPH Fluid Acceleration System V1.0. China Computer Software Copyright, Registration number: 2024SR0821036. (1st author)
  • Highlights:
    • Developed a software system enhancing computational speed by over 50%, enabling real-time fluid simulations in manufacturing (AI for Engineering).
    • Optimized for complex scenarios like metal casting and injection molding, leading to decreased material waste and improved product quality.
• J. Nan et al. (2020). A handling robot with an adjustable manipulator. Patent No. ZL 2019 2 1806945.7. (1st author)
  • Highlights:
    • Designed an adjustable manipulator with a 9-degree-of-freedom control system, significantly increasing flexibility and efficiency in industrial automation tasks.

📚 Papers Published in International Conferences

• J. Nan. (2023). Simulation study of axial ultrasonic vibration micro-milling of TC4 titanium alloy based on ABAQUS. ICCSMT 2023 (Sino-Germany), Oral Report & Letter.
  • Highlights:
    • Demonstrated that ultrasonic vibration significantly reduces cutting forces and heat generation, leading to improved tool life and surface finish.
• J. Nan. (2023). Design and Research on Heating System and Hot Bending Process of 3D Glass Bending Machine Using ABAQUS and Particle Swarm Optimization. ICCSMT 2023 (Sino-Germany), Oral Report & Letter.
  • Highlights:
    • Applied Particle Swarm Optimization (PSO) to refine the heating system’s power distribution, achieving improved temperature uniformity and bending accuracy.

🎼 Working Papers

• J. Xu, J. Nan, P. Feng, F. Feng. Physics-Informed Neural Networks for Burr Fracture Prediction in 3D Elastic Structures. (To be submitted)
  • Highlights:
    • Proposed a Physics-Informed Neural Network (PINN) that embeds physical constraints into the neural network, reducing inference time to <1s compared to 18s for FEM simulations.
• J. Nan, J. Xu, P. Feng, F. Feng. Graph Neural Network-Enhanced Chip Heat Dissipation Simulation for PCB Components with Multi-Phase Solids and Fluids. (To be submitted)
  • Highlights:
    • Utilized GNNs to simulate conjugate heat transfer for PCB components, achieving <1.2% prediction error while significantly reducing computational time.
• An Atomic Skill Library Construction Method Combined Embodiment VLA and VLP for Industrial Applications. (To be submitted)
  • Highlights:
    • Proposed a data-driven method using Vision-Language Models (VLP & VLA) to decompose complex industrial tasks into reusable atomic skills for robots.