Summary: | 碩士 === 朝陽大學 === 工業工程與管理研究所 === 86 === Due to high automation of manufacturing equipment, high
speed of production rates, and small and precise specifications
of basic components in IC package process, the inspection
operations have used computer vision techniques to inspect
process products. If parameter settings of vision inspection
machines are not appropriate, the inspection outputs will
easily result in two judgment errors, i.e. statistical type I
and type II errors. In this research, two kinds of frequent
defects, mark defect and lead defect, in IC package process
are the major concerns needed to be improved. A procedure of
combining designs of experiments and response surface techniques
is proposed to find the best parameter settings of vision
inspection machines for reduction of the judgment errors of the
two kinds of defects. Three response variables, type I error,
type II error, and expected loss cost are implemented in this
research.
We find the parameter settings of different response variables
can be determined according to a company's process capability.
The vision inspection machine can not only correctly inspect
products but also save the quality costs of the products.
Finally, a confirmation testing with the best parameter setting
is performed. The results show that the correct inspection
rates are improved up to 99.81% in mark defect, 71.16% in lead
defect, and 85.49% in both of them.
|