| In furious market competition, multi-variety strategy has become a commonchoice to match market requirement. This thesis analysis complex multi-varietymanufacturing system such as semiconductor packaging, and include features ofproduct, yield and operation. The problems in such manufacturing system areconcluded in this thesis. At first, author reviews the existing research achievement of multi-varietymanufacturing system management. Author reviews some related investigate field,such as statistical process control, data mining and yield management. Then authorintroduces two level quality management framework based on data mining, includingquality monitoring level and quality improvement level. The details of the frameworkare presented. This thesis introduces a multi-variety statistical process controlprocedure based on cluster analysis and ANOVA. Furthermore I extend univariatemethod to multivariate situation. In order to resolve quality improvement problem, Iintroduce a low yield dataset extracting method based on k-means clustering, andfault orientation method. These technologies and methodologies are integrated intoquality manufacturing framework. The relationship of yield and equipment is discussed in the dissertation inmulti-variety manufacturing environment. And author put forward the adaptability ofvariety and equipment in which this variety is produced can be performed in lot yieldlevel. So author introduces a two dimension dispatching idea include time and quality,and the dispatching method is described. Such idea extends quality management toshop dispatching level, and integration quality of manufacturing system can beimproved. Demonstration and simulation study are used to estimate available of newmethods. The result of demonstration is good enough. Through simulation study, theobject of improve system quality level can be achieved. In the end, a summary ismade, and some further investigation directions are pointed out.
|
PDF Full Text Request