Designing And Assembly Planning Of Bar System Based On Intelligent Methods

Bar system,which is widely used in aviation and space industries,has many advantages such as high specific strength,low manufacturing difficulty,on-site implementation and adjust feasibility.For these reasons,bar systems are often taken into consideration in advanced manufacturing techniques like light-weight design or flexible manufacturing.In pace with the improvement of the whole manufacturing industries,bar system has been more complicated than ever.This situation demands higher requirements in the design and assembly process of bar systems.However,different from the general mechanical system,bar systems have more complicated part-relationships.Meanwhile,their structures lack explicit patterns.For this reason,the design and assembly process have been bottlenecks of bar system manufacturing.Lacking patterns is an important fact of bar system designing and processing.This fact is reflected in topological framework designing,exact 3d modelling and sub-assembly grouping and so on.The development of bar system has to rely on existing work and human experience.For this reason,an import problem is to transfer the uncertain knowledge implied in designed models and processed data of bar systems to reusable,transferable explicit knowledge.With the superiority of intelligent method for uncertain problems,we studied the principle,methodology and key technology of knowledge mining for bar system designing and assembly processing.The main aspects studied are listed as follows.1.A structure design optimizing method of bar system for support forces distribution is proposed.The optimum object of support forces distribution is defined at first.The object is to minimize the deviation of actual support forces distribution and the ideal support forces distribution,where the deviation is expressed with a root mean squared value.To minimize this deviation,a gradient direction for adjusting the section area of bar parts is deduced from the static equilibrium equation in finite element form.This direction is added into a particle swarm algorithm to iteratively search the optimum support forces distribution of bar systems as requirements.Meanwhile,the affection of external forces added to bar systems is considered.The searching range of the exerting forces is deduced and added into the particle swarm algorithm to solve an optimum exerting forces scheme.2.An intelligent precise 3-D modelling method for bar system based on a union of deep-learning and Chu-Liu-Edmond algorithm is proposed.At first,the procedure of 3-D bar system modelling is analyzed and divided into overlapping detection stage and geometry calculation stage.The latter is proved to be a deterministic procedure.The uncertainty is limited in the first stage.Based on this conclusion,a deep-learning method is proposed to recognize the relationship between two parts at first.The recognitions of multi parts-pair consist of a weighted directed graph.Then the Chu-Liu-Edmond algorithm is used to eliminate inconsistent exist in the graph and finally generate a reasonable overlapping solution for bar system.3.An intelligent assembly liaison graph generating method for bar systems based on support vector machine is proposed.The difference between bar system and general mechanical system in assembly was analyzed.Against lacking geometrical matching features among parts in bar system,the relationship between part-pair is expressed via oriented boundary box(OBB).The part OBB,parts group OBB and parts intersection OBB is defined to represent assembly feature.To accelerate the calculation of defined OBBs,a fast iterative algorithm is introduced and its algorithm complexity and effectiveness are analyzed.Based on the defined OBBs,feature vectors are built to represent the relationship between two parts.By the vector data extracted from existing models,a support vector machine is trained.With this scheme,the knowledge of assembly relationship of welded bar system is obtained.4.An intelligent sub-assembly grouping method based on community detection is proposed.A criterion of sub-assembly of bar system is given.The assembly between two parts is weighted by a defined calculation based on OBB to represent the compactness of the assembly.The calculation of part assembly leads to an undirected weighted graph model.The Girvan–Newman(GN)algorithm is introduced to detect the communities in the graph.To accelerate this process,a node fusing algorithm is proposed to reduce the number of nodes and edges of the graph while keeping the remain edge betweenness unchanged.The detected communities are evaluated with given criteria iteratively to optimize the sub-assembly grouping of bar system.Taking design and assembly process of bar system as our research target,the optimum design of topological framework,precise 3-D modelling,assembly liaison graph generating and sub-assembly grouping of bar system is studied with the intelligent methods.Case studies showed the effectiveness of the proposed methods.