Research And Application Of Transferable Digital Twin-based Fault Diagnosis Model For Cranes

To ensure the stable and reliable operation of vehicles,high-fidelity digital twin models are used to achieve the true mapping of equipment.This high-fidelity model not only includes detailed geometric models but also includes internal models such as mechanism models and data models,requiring a high degree of information dimension.Therefore,modeling is difficult,and the implementation cycle is long,which cannot meet the timely mapping when vehicle equipment is replaced.Reusing existing digital twin models,achiev ing rapid model migration and utilization is becoming one of the hotspots in digital twin modeling.This article proposes a transferable modeling method for digital twin models to address how to reuse the multi-dimensional information of existing models to improve modeling efficiency.The main content includes :(1)In response to the issue of single vehicle fault diagnosis algorithms,difficulty in quickly modeling to address multiple devices and various fault types,a transferable digital twin modeling met hod for vehicle fault diagnosis is proposed.This method uses meta-models to construct digital twin assets and builds geometric,behavioral,and algorithmic dimensions of digital twin models through digital twin asset construction methods,realizing a unif ied mapping of physical entities in a fixed form.(2)In response to the complexity of vehicle operation mechanisms,when equipment is replaced or working conditions change,it is difficult to accurately model and adapt to various scenarios and working conditions.A digital twin multi-dimensional migration method is proposed.This method starts with the digital twin model’s multi-dimensional migration framework and formulates the migration process and strategy between the three dimensions,including achieving interoperability between models using geometric mod el assets;combining mechanism model assets and Io T data assets to achieve migration between behavioral dimensions;and transferring knowledge of the algorithm dimension based on transfer learning.(3)In response to the complexity and tediousness of const ructing vehicle digital twin models,and the lack of an implementation platform for unified model construction and migration,we have explored a transferable digital twin system building method based on digital twin assets.This method starts with the technical framework and overall framework,carries out detailed design at various levels,proposes various module design schemes,and discusses their functions.Finally,the system architecture design is carried out in conjunction with the above discussions.Finally,in response to the practical application needs of intelligent monitoring of speed reducers in the trolley part of the vehicle,a transferable digital twin system based on digital twin assets is developed.At the same time,the multi-dimensional migration method is verified,and the differences in the time required for the construction of digital twin models using transfer learning methods and traditional construction methods are compared and analyzed in different scenarios.The case shows that the m ethod proposed in this paper has good application value.