Investigation On Lightweight Technology Of Digital Twins Production Line CAD Models

In the digital twin production line scenario,to achieve accuracy and efficient virtual-real linkage between the virtual model and the physical equipment on site,the reliability,authenticity,and the real-time data of the virtual equipment 3D model are extremely demanded,which is static,dynamic and full lifecycle.However,the physical equipment of the production line is mainly described by the mechanical CAD assembly model,which retains a complex topology and rich geometric details.It has an extremely negative impact on the degree of smooth and accurate interaction between digital and physical entities.Because its unique features are failed to take into account by the mainstream conventional mesh simplification algorithm,a better simplification effect can not be achieved.At the same time,the scene of the digital twin production line is complex,and there is an urgent need for dynamic and efficient division management of the scene,which reduces the number of queries in the process of invisible model culling.To address the above issues,we propose a lightweight method for the scenes of digital twin production line models,in which the significant geometric features of the models can be effectively maintained while the model data is reduced and the running frame rate of the scenes is raised,and also dynamic real-time multi-scale lightweight utilizing parallelized computational acceleration is realized.Firstly,a lightweight method framework for digital twin production line model scenes is proposed.By analyzing the unique features of common mechanical CAD assembly models,a lightweight method framework for digital twin production line model scenes is constructed based on the needs of designers and operators of digital twin production lines and the characteristics of mechanical CAD models,and the technical implementation scheme is illustrated.Secondly,a QEM mesh simplification algorithm optimized for digital twin model features is proposed,and significant geometric features of the model are maintained.The redundant model data are subjected to a de-weighting pre-processing operation,constraints for discrete fold plane features and other common constraints are introduced,the best candidate collapsing vertex calculation strategy is designed,a penalty function is constructed by combining the triangle regularity,and finally,an optimized QEM mesh simplification algorithm is established.Then,an interactive marker-based interface identification and culling method are proposed to support the identification and culling of the invisible interfaces existing in the assembly model.The simplified model is pre-processed with homogenization and a hierarchical queue stack is constructed for data flow management,a modified watershed grid partitioning algorithm is executed,the over-segmented regions are merged utilizing the adjacency hierarchy connectivity map,and finally,the invisible interfaces are identified and simplified by combining interface screening rules and interactive user markers.Finally,scene division optimization and local adaptive culling methods are proposed to identify redundant model data that cannot produce interference on the screen space and make them culled.By constructing a hybrid data structure of KD-Tree and BVH,the digital twin scene is divided efficiently,and the adaptive culling operation of the redundant model data is realized in the focused local nodes after scene division according to the spatial occlusion of the model and the view cone interference relationship.Finally,a digital twin experimental environment is built and a CPU-GPU heterogeneous parallelized computing acceleration platform is constructed.In the digital twin experiment scenario,common mechanical CAD assembly models are utilized to perform experiments of mesh segmentation and interface identification,parallel lightweight scale partitioning performance comparison experiments,parallel lightweight acceleration performance comparison experiments,and model lightweight results experiments.Finally,a digital twin scenario of a smartphone assembly line is built to demonstrate the practical application effect after the lightweight method.The experimental results and demonstration results show that the high application value and research value of the lightweight method in this thesis are proved in the digital twin production line scenario.