• 목록

As profit of steel enterprises is getting smaller and comparison is getting harder, in steel manufacturing industry quality inspection and control have become main measures for getting over the difficulties. In sequential manufacturing process, failure occurred in preceding process make final quality worse. Detecting the factor which affects to final quality is necessary to make process stable and efficient. Existing studies to improve the surface quality of small steel bars assume that these quality factor have been identified and focus on diagnosis of defects. However, there are no attempts to validate the quality factor based on data. In this paper, we attempt to verify the quality factor based on data using statistical and data-mining techniques. To get over practical problems from variation of quality due to operation date and merge measurement of quality, we suggest method using statistical significance difference and using k-means clustering(k=2). Method using statistical significance difference considers the overall tendency about quality and method using k-means clustering consider the tendency toward outlier. Quality factors of small steel bar can be detected by using both methods serially. And we apply this method to real world data in case study.