Research On Technology Of Knowledge Discovery For Process Planning Information System

Process knowledge is an important enterprise knowledge resource in product designing and manufacturing and an indispensable part of information technology system in manufacturing industries. It is also the core and key element in realizing the intelligence of process planning information system. However, owing to the diversity, complexity, experience, uncertainness, etc., and other characteristics of process knowledge, the process knowledge in enterprises is in need of effective generalization and arrangement, which makes it hard to be reused and shared. How to formalize and digitalize the process knowledge so as to preserve them in explicit forms is one of the front studies in intelligence of process planning information system.With the progress of enterprise informationization, Computer Aided Process Planning (CAPP) system has been widely applied in manufacturing industries, which has accumulated a large quantity of process planning data in forms of Word, Excel, Cards and other electronic documents. These planning data is the crystallization of the experts and contains a vast of process knowledge. How to discover, acquire and reuse the implied process knowledge from these various-structured data is an important and urgent problem in the intelligence of process planning information system. Supported by the project of national natural science fund and by adopting Data Mining theories, this research focuses upon the basic principles of process knowledge description, classification, discovery and reuse in the orientation of process planning information system, and the implementation arithmetic in computer network environment of these principles, so as to explore a technique to accumulate process knowledge and increase intelligence level of process planning informationsecurity management model in the orientation of user, role and task is presented. The architectural structure of knowledge-based process planning information system is further established. In order to support the discovery of process knowledge, a part process information model is set up, which is an integration of such features as management, geometry, material, precision, assembly and process.(2) A process knowledge classification method is presented based on Data Mining technique. Process knowledge is classified into reusable process knowledge and decisional process knowledge. The former is composed of standard process term, process unit, process template and process case, the latter is constituted of selection rules for processing means, process equipment, process parameters and man-hour ration, process routes arrangement, etc. According to the structural characteristics of these two kinds of process knowledge, the relational databases storage models are defined respectively and the functional models of process knowledge repository administration system is constructed which integrates knowledge-collecting, management and service.(3) An improved hierarchical agglomerative method is designed to make a clustering analysis of the process data according to the features of process data. On this basis, a standard sentence-acquisition method is put forward through editing distance measurement of similarities between different classifications; A double-level process document model is constructed, which is composed of process procedure and steps, to measure the similarities of process documents and carry out clustering analysis on process data so as to acquire process cases.(4) The problem of knowledge discovery in sequence pattern is discussed, and process document is simplified into process sequence, and then a formalized definition is given. By combining Apriori arithmetic and the processing sequence in working procedure, Apriori-seq arithmetic is designed to mine the process sequence pattern so as to acquire process unit and process template.(5) To store the process decision data samples by Process Decision Data Table (PDDT). Designed an improved process decision tree studying arithmetic based on CLS (Concept Learning System). Measuring the importance of conditional attribute for process decision rule in PDDT by ID3 (Iterative Dichotomizer 3) algorithm. Accordingly a process decision tree is established, the principle of simplified disposal of process decision tree is put forward and the method of transformation from process decision tree to process decision regulations is presented.(6) A Directed Assembly Relation Graph (DARG) is defined orienting process cases searching. A case searching algorithm based on DARG is designed for parts with complex assembly restrictions. A Formal Part Query Language (FPQL) is defined, based on which a speedy process case searching algorithm is put forward for parts with simple assembly restrictions.(7) A general goal and functional model are put forward in accordance with the development and requirement of process planning information system in JiaJiang Hydraulic Engineering Machinery Factory (SGCAPP). By adopting process data formalization, and based upon XMLlocal data exchange and Plug-in, a software system is developed which is composed of project management module, flow management module, resource management module and process planning module. To realize process knowledge administration of process unit, typical process and process case, etc. in process resource management module and thus the separation of knowledge administration and application is realized. SGCAPP system has achieved favorable results in enterprise practices and was checked and accepted in April 2005.