Research On Multi-mode Case Retrieval And Evaluation For MEMS Product Conceptual Design

With widely used in many domains, for example, automotive electronics, aerospace,information communication, biological medicine, automatic control, national defense andmilitary, MEMS has a considerable market prospect in the near future. Because of the lack ofmature MEMS design theory and method, the key to MEMS CAD is how to reuse priordesign knowledge and experience in order to support MEMS conceptual design. For thepolymorphism and dynamics of MEMS design knowledge and the freedom of conceptualdesign, an approach of multi-mode case retrieval and evaluation for MEMS conceptualdesign is proposed to share and reuse MEMS design knowledge, improve conceptual designefficiency and effectiveness. The main contributions and conclusions of the present work aresummarized as follows:A framework of multi-mode case retrieval and evaluation for MEMS conceptual design iscreated. In this framework, MEMS designers can represent design problems as functional,parameter and structural information, and select such case retrieval methods as functionalretrieval, parameter retrieval and structural retrieval freely to achieve flexible MEMS caseretrieval, and finally carry out a fuzzy evaluation of the retrieval results.A multi-view representation of MEMS case is proposed. The definition of MEMS caseontology is analyzed, and then the MEMS case model including functional, parameter andstructural information is created and represented by the object oriented method. At the sametime a hierarchic case library is also created.The methods for MEMS case functional retrieval, parameter retrieval and structuralretrieval are proposed respectively. The evaluation model of functional similarity and casequality is firstly created based on weighted keywords and cited frequency, and then themulti-keyword functional retrieval method is implemented based on double-keyword retrieval.Secondly, the objective weights are computed by optimized entropy model and the subjectiveones are computed by group decision making model, and then parameter case retrieval isachieved by the nearest neighbor approach. At last, the generalized and special structuralconcepts of MEMS product are respectively defined. Generalized structural similarity is computed by directed graph model and its permanent function, while special structuralsimilarity is computed by the edit distance between path sets of MEMS product trees, whichis based on the computation of the edit distance between double paths.The methods for MEMS case fuzzy evaluation based on normal fuzzy set and two-tuplelinguistic information are proposed respectively. The evaluation architecture and criteria arefirstly created. Then, the similarity is computed based on normal fuzzy set. The casequantitative evaluation based on normal fuzzy set is implemented by combining TOPSIS. Atlast, an optimized two-tuple linguistic model is created by the combination of linear scale andexponential scale, and the method for integrating multi-agent evaluation information withmulti-granularity is proposed. Then the qualitative evaluation method for multi-granular andtwo-tuple linguistic information for multi-agent is studied.Finally, a software prototype is developed based on the above researches by Delphi7andSQL Server2000. And the architecture, work process and main modules are analyzedrespectively. Then the effectiveness of the system is demonstrated by filter retrieval.The dedicative works presented not only achieve the share and reuse of MEMS designknowledge, and help designers to work efficiently and effectively, but also create versatileand promising pathways for case retrieval, promote the application of CBR to MEMS designand the development of MEMS CAD, and indicate theoretical and practical significances.