An adaptive digital twin-based method for resilient design: a case study of high-speed train bogie

With the increasing diversity and complexity of industrial products, product design must focus on its resilience to enhance stability in response to potential external influences. Digital twin (DT) technology enables mapping and updating between virtual models and physical systems, improving the accuracy of design results. However, when the environment changes, current DT design method typically requires re-solving based on the updated model. This may lead to loss of design information, interruption of the optimisation process and longer cycles, reducing the design continuity and response efficiency of the system. In order to address the above issues, this paper proposes an adaptive digital twin-based method for resilient design. Firstly, a gradient boosted regression tree (GBRT) is used to construct mapping relationships between design parameters and multidimensional objectives for complex products, and realises the prediction and updating of non-linear relationships. Then, a resilient optimisation strategy based on DT is proposed. The strategy evaluates the relevance of the model update process and defines a dynamic non-dominated sorting mechanism for fast adaptive iteration of the search population, which improves the reusability of the model and the efficiency of the optimisation design. Finally, a case study of the high-speed train bogie is presented.