Research On Knowledge Representation And Solving Algorithm Of Product Configuration

Configuration problems have become a very active research field, whichis an important application of AI techniques developing with the productmode of mass customization. As a branch of theory research of AI techniques,configuration problems have attracted much attention by many researchers ofthe application field these ten years.It needs two steps to solve a configuration task. The first step is torepresent domain knowledge of special area. The second is to conduct acorresponding algorithm to solve the problem based on this way ofknowledge representation.Recently there are many different methods of knowledge representationof configuration. This paper is focus on two main methods: rule-based andCSP–based. By mean of comparing these methods and considering the realsituation of product management of manufacture field, we propose aGraph-based knowledge representation method. The structure of a product ora part is usually represented by BOM(bill of materials) in enterprise, whichcan only reflect the structure and quantity between a product and thecomposed parts. We can also make use of BOM to design or arrange newproducts. However, BOM is fixed for a product,which can not reflect domainknowledge of product family and order the product satisfying usersrequirement flexible. Studying the way of knowledge representation ofproduct family, we improve the management pattern of BOM. Combining thetechniques of constraint reasoning ,the static structure of BOM can bechange to a virtual and dynamic general product structure by mean ofadding parameters into component nodes of a product. General productstructure tree can represent a serial product and be adapted to different kindsof knowledge representation of product configuration, which also can reflectthe manage pattern of product configuration and build the connection andconstraint between component and their attributes or components themselves.By use of general product structure tree, it can make use of search algorithmspresented and can realize incorporating the system of configuration andPDM.Explanation generation has a long history and became more and moreimportant these years ago in expert system. During the solving aconfiguration,one often face the problems like this: which chooses productconflict?why my choice make no solution? Why my favorite value cannot selected? .etc. In order to answer the questions, configuration systemshould provide useful messengers to tell user the reason and guild them findthe products they need. This paper describes the definition of explanation andclassifies it into three types: elimination, entailment and contradiction.Because of elimination recording the reason of deleted value, we can makeuse of it to guild backtrack. We give color map problem to illustrate how tosolve explanation.Configuration problem can be considered as a user-based CSP. Becauseof powerful knowledge expression and many algorithm presented, we canconvert a configuration problem into a CSP. There are many algorithms tosolving CSP. Backtrack always think that the last node is the reason the resentnode failed, but in many situations there is no relation between the resentnode and the last one. Backjumping and dynamic backtracking can skip thenode that has no impact on resent node and go back to the node who deletesvalue directly. We propose a solving algorithm of configuration by use of theidea of dynamic backtracking. In general product structure tree, a noderepresents a component, so the root node can be seen as the product. Theprocess of solving a product is to ensure validity of the component denotedby the root node. As for an unfixed component, we can start from its functionand structure. We first fix its function and then its structure. It is obviouslythat solving a product is a recursive process.In the face of requirement changing quickly, designers must think outdifferent kinds of new products to adapt to this situation. In fact, in order toimprove the quality and efficiency of product design, most new products aredeveloped on the basis of some existing similar products. Therefore,designing a product often can be seen as redesign. In a certain extent, theideal product redesign is the key to success of enterprise. Product redesignhas become a problem faced by enterprises. There are many experts havingbeen done research for this problem and presented some methods to solving it.We have described the application to product design based on model-baseddiagnosis and presented the concepts of redesign diagnosis. In additional, wehave developed a problem solving algorithm for redesign based on diagnosis.