The Study On The Construction Method Of Design-Oriented And Evaluation-Oriented Expert System For Relay

As a core of Intelligent CAD, the design-oriented expert system aimed at automation and intelligent designing has been quick developing although there still are some problems to be discussed solved. Study on the methods of knowledge model establishing and knowledge processing and on the designing for system under conditions of partial and uncertain design information consist of the main part of this work. Also studied is on application technology of optimizing and comprehensive evaluating for products' design.This work takes the relay products as the object to study, the flowing three aspects of product designing, optimizing and evaluating are the main part of this work:(1) To the question of the methods for representing design procedures and objects, put forward Design Mode-Oriented model (DMOM) and three kinds of design object models (conceptual model) matched with DMOM: Functional model, Parametric model and Case model.(2) In the knowledge utilizing, based on the above conceptual models, three modes are given as Functional mode, Parametric mode and Case mode. The system inference methods and control strategies based on the design modes are studied to solve the design problems with different design modes according to users' design requirements.(3) For Parametric model, put forward the conception of Feature model, as well as two fuzzy algorithms and a hybrid algorithm to design products under the guide of Feature models based on the closeness and membership degree is given. At the same time, conflict resolution strategy for rule set and fuzzy decision method with product design parameters subjected to range.(4) For Case model, it gives the definition to the similarity degree between the design problem and case and the method to index and match cases based on the comprehensive consideration of various design characteristics of products. This method is suitable for CBR when functional, behavioral, structural and parametric attributes or fuzzy and exact design parameters coexist in design problem.(5) In machine learning, Clustering analysis/Case-base reasoning- machine learning- fuzzy recognition(C/CMR) is put forward to determine the Parametric model or implement the structure design during the period of the system developing or utilizing. In addition, a new method to distributing weights to products parametric attributes is given based on the level dividing of products types.(6) The using of hybrid Genetic Algorithm (HGA), which hybridizes Genetic Algorithm and Simplex algorithm is to study the optimizing method to design products.(7) Multi-objects and multi-level index system is established for evaluating the product's design scheme. The method of establishing Standard models which are representative of different grades product by clustering analysis and the validity of clustering are discussed. At the same time, by using the evaluation method based on the fuzzy recognition theory, the two level mapping of from objects evaluated to Standard model and from Standard model to fuzzy comment is set up and fuzzy transform induced from fuzzy mapping rules and relative mapping methods are given too.