Fault data augmentation for electromechanical coupling systems based on hierarchical collaborative digital twin: System data augmentation based on digital twin

Effective fault diagnosis is crucial for the reliable running of Electromechanical coupling Systems (EMS), yet hampered by insufficient entity fault data. Digital Twin (DT) technology offers the potential for virtual fault data augmentation and fault diagnosis improvement. However, there is still a lack of an effective and systematic methodology, to decouple complicated EMS entities and construct their full-system DT. To address this, a hierarchical collaborative DT construction framework is proposed for fault data augmentation of EMS. Specially, we decouple EMS entity into the triplet representations of element, data, and relationship, which establish the profound understanding of coupling characteristics from multiple modalities. Furthermore, we develop a hierarchical DT modeling method to mirror these complicated couplings as four-level sub-DTs of space, behavior, process, and status. Each level of sub-DT utilizes the data-mechanism combined technique to balance modeling adaptability and precision. Finally, these heterogeneous sub-DTs are integrated as full-system DT driven by collaborative orchestration algorithm, which achieves the global consistency mirror with the real fault manifestation under diverse fault modes. Experiments on a multi-coupled electromechanical fault test bench validate our framework. Results exhibit the average improvements of 17.29 % and 9.97 % in accuracy of data augmentation fault classification, confirming its superiority and effectiveness.