Research And Application Of Throughput Model Based On Data Warehouse And Case-based Reasoning

Data warehouse technology has become one of the research hotspots today. As a kind of optimum management, enterprise's data solution that offers decision supports, data warehouse technology has already been adopted and functioned by a lot of enterprises.Case-based reasoning is a kind of logical descriptions that used to solve realistic problems. Comparing to traditional rule-based system, it has many advantages, which can predigest knowledge gathering and accumulate knowledge easily.Based on each machine production processes of Shanghai BaoSteel CO, LTD. cold rolling thin strip mill, combined with the development of optimizing production planning and production scheduling system in the mill, this paper studied model building technology that used to predict machine throughput, and has build machine throughput model based on data warehouse and case-based reasoning.In this paper, data driven method has been adopted to design data warehouse: By using the existed data that belong to the legacy database of the enterprise, the data and the relationships of the data have been reconsidered according to the analyzing fields, the subjects in the data warehouse have been organized, the "commonalities" of the data and subject data in the data warehouse has been discerned effectively by using the data model, and the attributes which connect each other among the subjects have been set up. This model uses data mining to analyze the production conditions of the whole cold rolling thin strip production processes, and builds local data warehouse (independent data mart) that used by production planning department and production administration department, using historical production data of the enterprise, cooperating with the on-site producer's experience, builds case base for the case-based reasoning method, thus realizes taking historical production data and expert experience as basis to predict the machine throughput.This model can predict each machine throughput accurately. This model has the capacity of self-learning and self-adjustability; it can reflect the throughput changes due to such factors as technology improving and the varieties of processing changes automatically. It has strong adaptability.