Visualization and interaction design methods for digital twin systems in smart manufacturing
This paper proposes an advanced 5D digital twin framework specifically designed for intelligent manufacturing scenarios with nonlinear, high-dimensional process dynamic characteristics. By systematically integrating geometric, temporal, physical, behavioral, and probabilistic dimensions, the proposed system extends the traditional digital twin paradigm, enabling it to comprehensively and accurately model complex manufacturing environments. Embedding physical priors and real-time probabilistic reasoning, as well as cross-domain data fusion, adaptive data preprocessing, and high-frequency industrial sensor networks—particularly LSTM-based simulation modules—constitute this architecture. To achieve interaction and monitoring, a modular visualization interface was developed. Supports dynamic process analysis, anomaly localization, and intuitive operator involvement. In the actual thermal management tasks of battery production lines, experimental validation shows that compared to traditional digital twin configurations, the 5D method improves prediction accuracy, shortens anomaly detection intervals, and enhances operational usability. This framework provides quantifiable performance metrics for benchmarking complex system behavior and supports multimodal input streams. The results highlight the 5D framework's capabilities in achieving smart manufacturing, including unifying heterogeneous industrial data sources, driving high-fidelity simulations, and providing actionable visual analytics. This comprehensive strategy makes the system more transparent, scalable, and robust.