Research Of Assembly Sequence Planning Based On Knowledge And Its Key Problems

Recently years, academia has paid more attention to advanced manufacturing technologies, such as virtual manufacturing and agile manufacturing. It becomes a hot research field, and a lot of progress was obtained. Assembly Sequence Planning(ASP) is the core of the Computer Aided Assebmly Planning(CAPP), which has attracted attention widely.An assembly is composed of parts, which can come into being a product in different order, so different sequences are formed. The sequences have different affects. Some of them can build the product successfully, and other of them are defeated. The aim of the ASP is to find the best sequence from the sequences. That is to say, ASP belongs to the optimized problem, which should satisfy geometry, physics, and machine constrains, and so on.Advantages and disadvantages are in current ASP at the same time. ASP which is based on geometry's reasoning is universal, but its time-consuming is unavoidably. Igoring non-geometry information in ASP, the sequences are unreasonable. So based on the review of existing research works, this dissertation researches ASP based on knowledge.Designing and setting up of the knowledge library is crucial in the system of ASP based on knowledge, and they are the foundations of assembly model and assembly planning in the future. According to the researchment of assembly's knowledge, three kinds of knowledge are set up. The first class is machine's parts which are gotten together by designers. The second is connecter-structure knowledge which reacted strongly to product assembly. And the third is rules knowledge. According to the three kinds of knowledge, parts library of machines, case base, and assembly rules system are built, and they are base of the ASP.Assembly model is the foundation of the assembly sequence planning. Based on the researchement of assembly model, this dissertation creates Assembly Petri Net Model(APNM) to represent the assembly relation of parts. In AKNM, places denote parts and transitions denote assembly operations. Connect-relations of the parts are denoted by relation of the places and transitions. In order to express the relation rather, assembly knowledge is introduced. That is to say, attributions of the parts are knowledge places and assembly rules are knowledge transitions. This two kinds of knowledge are added to APNM, Assembly Knowledge based Petri Net Model (AKBPNM) is formed. For the simple assembly, AKBPNM is enough to get its assembly planning sequences. While the assembly's structure is complicated, it is unavoidable that combinatorial explosion occur in getting assembly sequences by AKBPNM. In order to avoid that matter, the connector-structure is used to simplify the assembly model. The dissertation puts forword that consistency must be between the simplified model and the original model. And the attribute is checked by related-matrix. At last, the assembly sequence is gotten by the related-matrix.Many feasible assembly sequences exist in an assembly, but how to confirm a sequence is fine or not is an import problem. So a method must be find to evaluate these sequences. On the base of researchment of the existent evaluations, the evaluation of mechanical product assembly based on assembly operation is put forward. Degree of assembly operation, technologic effection of assembly operation, and tool of assembly operation are taken into account while the sequences are evaluated.Then, this dissertation gives a review of Ant Colony Algorithm, and based on a dissective analysis of its slow searching speed and premature convergence, a new Ant Colony Optimization(ACO) algorithm is proposed, in which three aspects is improved. (1) Information of assembly operation is added to the state transfer function. The allowed transfer area is confirmed by the assembly relation of the parts. (2) In order to speed up convergence of the Ant Colony Algorithm, and avoid the loacal optimization, the mix updating rule is used to update pheromone. During evolution, if coefficient of pheromone IS changed along with time, the better solution is obtained. (3) Via analyzing the affects of the parameters in the state transfer function, the better solution is obtained while change the fixed parameters to transforming with time. Through theoretical analysis and experiment, parameters'change can get the better solution. At last, by theoretical analysis, the convergence of the new method is proved. To improve the performance of the ant colony algorithm in solving large-scale assembly planning problem, scale compression thought is put forword. The algorithm based on problem scale compression and solution scale compression. Finially, the validity is proved by an experiment in solving the large-scale assembly sequence planning.Finally, this dissertaion developed an Assembly Planning and Analyzing System with UG/OPEN API and VC6.0, in the environment of Unigraphics NX.