Research On Optimization Methods Of Product Design Structure Clustering

The hierarchy clustering of product design structure is an important part of theproduct design module. In this paper, design structure matrix is researched as the basisfor the conventional method, and the path search algorithm is used to optimize for DSMclustering with the graph theory and the matrix theory as the theoretical basis. Thismethod can achieve hierarchy clustering of product design structure. However, thedigitized DSM models cannot be optimized by the path search algorithm, so intelligentalgorithm is explored to the self-organizing feature map. Corresponding improvementswere made to meet the division level clustering of product design structure andapplication examples are verified. The specific contents are as follows:Firstly, the concepts related to product design structure and the process modeling arefocused on introduceing. The tree diagram, directed graph, matrix are applyied to depictthe product design structure, and their mutual conversion between the three are alsodescribed. The clustering model of product design structure and the clustercomprehensive evaluation are analysised.Secondly, the DSM models based on the relationship between the elements ofproduct design are established.The DSM clustering optimization algorithm is morecomplex computing and the result of clustering division of complex products is notsatisfactory. Using path search algorithm to improve its. The model of product designprocess is compiled by C++language,which achieves on automation calculating withcomputer. The QTZ80H tower crane boom is used as an example to illustrate the processmodel. The results show that this method can effectively be used.Then, the path search algorithms can not divide clustering about digitized DSMmodels. By studying intelligent algorithms operating principle of self-organizing featuremap SOM method, the improved method can meet the DSM model in digital productdesign structure based clustering division.The level partitioning algorithm is added todivide block-level clustering of product design structure. Application of MATLABprogramming combined with neural network toolbox to achieve the method ofcomputing and clustering results visualization. And the method that is verified byexamples of Boolean DSM (0-1matrix) model or numeric DSM model established between product design elements are feasible to use level clustering division.Finally,the clustering hierarchy division results of product design structure based onthe original DSM method, the path search algorithm, SONN method and SOM methodare compared from two examples of auto body and motorcycle engine. And the clusteringresults are evaluated by the cluster comprehensive evaluation. The results show that theimproved of the SOM method is better effect on the hierarchy clustering of productdesign structure.