| Multidisciplinary coupled high-precision products such as electro-hydraulic servo valves,industrial robots and aero-engines belong to the "bottleneck" type of products in China,and their assembly-commissioning process has the characteristics of high complexity,high precision and high dynamics.However,the existing assembly technology requires a large proportion of manual experience and operation tasks,and usually requires repeated manual assembly-commissioning to meet the design requirements of high-precision products in terms of geometric accuracy and performance accuracy at the same time.With the continuous improvement of the complexity and precision of high-precision products,the existing assembly technology has been difficult to meet its market demand.In recent years,with the rapid development of technologies such as artificial intelligence,digital modeling and intelligent data collection,digital twins have been deeply studied.The digital twin realizes the two-way interaction and iterative optimization of the virtual and physical space by establishing a digital model equivalent to the physical object and dynamically evolving in the virtual space.To solve the problems of poor assembly quality and low assembly efficiency of current highprecision products,this paper proposes a theoretical research and application method of assembly-commissioning driven by digital twins.In the theoretical research method,the connotation of the digital twin installation and commissioning technology is revealed from the perspective of model and data;in the application process,the installation and commissioning quality control method and human-machine collaborative operation method under the digital twin environment are discussed.The main research contents of this paper include the following aspects:(1)Aiming at the problem of low fidelity and completeness of the digital model of high-precision products,a high-precision product digital modeling method for digital twin assembly is proposed.The formal expression of the high-precision product digital model is defined,and the product digital model including the virtual mapping model and the quality control model is constructed.The virtual mapping model integrates the MBD(Model Based Definition)integrated information model and the point cloud inverse model based on key features,which can evolve synchronously with the physical assembly process.The quality control model can make real-time predictions and analysis decisions for key quality nodes in the process of assembly-commissioning.(2)Aiming at the multi-source heterogeneity and multi-semantic heterogeneity of high-precision product digital twin installation data,a data fusion and knowledge generation method for digital twin installation is proposed.A multi-level ontology model of high-precision product digital twin assembly is constructed.In the MBD integrated information model,the fitting object entity and fitting information entity are extracted,and the mapping of the entity to the multi-level ontology model is realized.The multi-source heterogeneous virtual and real data are fused,and the multi-level ontology model is combined to build a knowledge map of digital twin installation and commissioning,which realizes the generation of digital twin installation and commissioning knowledge.In addition,the dynamic update of knowledge is realized by using the continuously accumulated data of the actual installation and commissioning process.(3)Aiming at the problem that manual experience is difficult to control the quality of assembly-commissioning in real time and accurately,a twin data-driven dynamic closed-loop control method for assembly-commissioning quality is proposed,and a dynamic closed-loop control framework for assembly-commissioning quality based on digital twin and cloud edge computing is constructed.At the geometric precision control level,a prediction method of assembly geometric error based on the state space model is proposed,and the error compensation scheme can be pushed to the physical assembly space by using the digital twin assembly knowledge map.In the aspect of performance precision control,a performance prediction method based on IETr Adaboost is proposed,which solves the problem of building a prediction model under high-dimensional and small sample data of high-precision products.The coupling performance optimization method realizes the qualitative guidance of the performance debugging process.(4)Aiming at the problem of low installation and commissioning efficiency caused by too many manual operation tasks,a human-machine cooperative installation and commissioning operation method based on digital twin technology is proposed,and a technical framework for realizing this method is constructed.At the task assignment level of human-machine collaboration,the division rules of the human-machine collaboration work area are analyzed,and a task assignment method for humanmachine collaboration based on capability assessment is proposed.At the level of the human-robot collaborative deployment strategy,the task sequence planning of humanrobot collaborative deployment based on digital twin deployment knowledge map is studied,and a robot deployment path planning method based on double-depth deterministic policy gradient algorithm is proposed,which ensures Rapid self-adaptive commissioning of human-robot collaboration strategy.Finally,this paper conducts a comprehensive production experimental verification of the digital twin assembly-commissioning theory.According to the research content,a digital twin installation and commissioning operating system is mainly designed and developed,which mainly includes a product digital modeling module,a data fusion and knowledge generation module,a quality dynamic closed-loop control module,and a human-machine cooperative installation and commissioning operation module.At the same time,taking the installation and commissioning of a certain type of high-precision servo valve as an example,the feasibility and effectiveness of the method proposed in this paper for solving engineering problems are verified. |
PDF Full Text Request