Research On Reduction And Knowledge Discovery From Uncertain Information On Floor Shop

As the manufacturing center of an enterprise, the shop floor must agilely operate to adapt to the modern manufacturing mode. Manufacturing Execution System (MES) are gradually taking their place in the world of Agile Manufacturing in last decade. This paper carries out a thorough study on some key issues of MES, which focus on the information processing technology for the optimization manufacturing processes on shop floor in the practical uncertain manufacturing environments. The systematic theories and methods are proposed for the uncertain information reduction and knowledge discovery from producing processes on floor shop.The correct scheduling and decisions in MES are based on the accurate information from the shop floor. In order to deal with the uncertain information from producing processes on shop floor and discover the useful knowledge for decision-making from uncertainty information, the method of uncertain information exploiting and knowledge discovery in databases (KDD) is investigated, and a solution based on the extension of Rough Set Theory, Dominance-based Rough Set Approach (DRSA), is put forward. The DRSA can extract both syntactic and semantic information from the stored data on shop floor, therefore, it can meet the requirements of uncertain information reduction and knowledge discovery in MES.By applying the DRSA, the uncertain factors on shop floor can be analyzed and reduced, consequently the uncertain information system can be compressed. A new reducts generation method, forward selection/backtracking algorithm (FS/BA), is introduced in this paper. The FS/BA not only can obtain all the exact reducts from the information system, but also can be extended to generate reducts in a more universal Rough Set Theory, Variable Consistency model of Dominance based Rough Set Approach (VC-DRSA). The extended FS/BA can be used for the probability analysis and uncertainty reduction for the information system. The fuzzy measure is introduced to evaluate the importance of the uncertain factors for decision-making on shop floor. By the fuzzy measure, the interdependence of two uncertain factors is computed, and best reduct is obtained. Since the FS/BA generates the reducts in a recursive way, the computation complexity of the algorithm is considerable. According to the definition of reduct, dominance matrix is constructed on the basis of dominance relation. This method is very effective in generating all exact reducts, due to its similarity to the discernibility matrix in the classical Rough Set Theory. The advantages and disadvantages of these two reduct generation algorithms are compared.The KDD from producing processes on shop floor is the non-trivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data. The objective of uncertain information analysis is to provide decision-making knowledge for the production planning and scheduling on shop floor. Based on the DRSA, the KDD methods are recommended for uncertain information on shop floor. An exhaustive robust rules set generation algorithm, which is based on the dominating/dominated local reducts, is presented. Based on the DOMLEM, the minimum non-robust rules set generation algorithm is developed; furthermore, the algorithm is extended to combine the probability information with rules set.The due date of a product is determined by the finishing time of the product on shop floor, which is an important task for the production planning and scheduling on shop floor. The determination of the finishing time is a complex problem affected by many resources constraints. A lot of researchers have devoted to to this problem and developed some models and approaches. However, most of these models are very complex, they need a large amount of data, and they do not consider the uncertain and unknown factors existed on shop floor. Therefore, these approaches are not effective in practice. Through the intensive investigation in a motor assembly shop floor, 9 kinds of uncertain factors which affected the due dates of the products are sorted out. An uncertain information system is constructed by the data collection in terms of the uncertain factors and product due dates from the shop floor. The uncertain factors are analyzed and reduced, and the useful knowledge of the decision-making rules is discovered from the information system by the KDD methods proposed in this thesis. The knowledge can be used for the due date prediction and uncertain factor analysis.